{ "cells": [ { "cell_type": "markdown", "id": "a5b2d48a-bce2-4f9e-bc77-a99d1f7cbd26", "metadata": {}, "source": [ "## 4 - The art of prediction\n", "### Asking scientific questions of models" ] }, { "cell_type": "markdown", "id": "85d10d91-f701-4ef5-bd1e-821fd5a81b22", "metadata": { "editable": true, "slideshow": { "slide_type": "" }, "tags": [] }, "source": [ "This week we have a very important topic to cover - using a model to generate predictions, and for us to test our scientific theories on those model-based predictions. \n", "\n", "This is likely one of the most important sessions we will cover, as it provides the basis for how to interact with models and have them answer our questons. We will cover:\n", "- Generating conditional predictions from a model\n", "- Generating marginal predictions and testing comparisons and slopes of those values\n", "- Translating our natural-language queries into statistical terms\n" ] }, { "cell_type": "markdown", "id": "64b24728-d557-44d0-9f15-c3d0d89df9c9", "metadata": { "editable": true, "slideshow": { "slide_type": "" }, "tags": [] }, "source": [ "### The `marginaleffects` package\n", "Much of the heavy lifting of our foray into predictions will be handled by a new package, called `marginaleffects`. This is a new Python package and can take a little getting used to, but it is very powerful. We'll explore how to use it and which of its functions we use to turn our questions into answers.\n", "\n", "Lets import everything we need, including `marginaleffects`, which we will call `me`." ] }, { "cell_type": "code", "execution_count": 1, "id": "c234ef3e-f825-48cb-bec2-3e065f169649", "metadata": { "editable": true, "slideshow": { "slide_type": "" }, "tags": [] }, "outputs": [], "source": [ "# Import most of what we need\n", "import pandas as pd # dataframes\n", "import seaborn as sns # plots\n", "import statsmodels.formula.api as smf # Models\n", "import marginaleffects as me # marginal effects\n", "\n", "# Set the style for plots\n", "sns.set_style('whitegrid') # a different theme!\n", "sns.set_context('talk')" ] }, { "cell_type": "markdown", "id": "372c0aa7-e196-4f43-a414-5e13079ac378", "metadata": { "editable": true, "slideshow": { "slide_type": "" }, "tags": [] }, "source": [ "### Fitting a model\n", "We will work through some examples with the `affairs` dataset, fitting models of varying complexity and using conditional and marginal predictions to understand the implications and answer our questions.\n", "\n", "First, let's load in the dataset, and use `statsmodels` to fit a simple model, akin to a t-test, which looks at whether those with and without children have higher marital satisfaction:" ] }, { "cell_type": "code", "execution_count": 2, "id": "8fa06949-f4be-4cd2-90cd-f03326b5e1e3", "metadata": {}, "outputs": [ { "data": { "text/html": [ "\n", "\n", "\n", " \n", "\n", "\n", " \n", "\n", "\n", " \n", "\n", "\n", " \n", "\n", "
OLS Regression Results
Dep. Variable: rating R-squared: 0.039
Model: OLS Adj. R-squared: 0.037
No. Observations: 601 F-statistic: 24.00
Covariance Type: nonrobust Prob (F-statistic): 1.24e-06
\n", "\n", "\n", " \n", "\n", "\n", " \n", "\n", "\n", " \n", "\n", "
coef std err t P>|t| [0.025 0.975]
Intercept 4.2749 0.083 51.635 0.000 4.112 4.437
children[T.yes] -0.4795 0.098 -4.899 0.000 -0.672 -0.287


Notes:
[1] Standard Errors assume that the covariance matrix of the errors is correctly specified." ], "text/latex": [ "\\begin{center}\n", "\\begin{tabular}{lclc}\n", "\\toprule\n", "\\textbf{Dep. Variable:} & rating & \\textbf{ R-squared: } & 0.039 \\\\\n", "\\textbf{Model:} & OLS & \\textbf{ Adj. R-squared: } & 0.037 \\\\\n", "\\textbf{No. Observations:} & 601 & \\textbf{ F-statistic: } & 24.00 \\\\\n", "\\textbf{Covariance Type:} & nonrobust & \\textbf{ Prob (F-statistic):} & 1.24e-06 \\\\\n", "\\bottomrule\n", "\\end{tabular}\n", "\\begin{tabular}{lcccccc}\n", " & \\textbf{coef} & \\textbf{std err} & \\textbf{t} & \\textbf{P$> |$t$|$} & \\textbf{[0.025} & \\textbf{0.975]} \\\\\n", "\\midrule\n", "\\textbf{Intercept} & 4.2749 & 0.083 & 51.635 & 0.000 & 4.112 & 4.437 \\\\\n", "\\textbf{children[T.yes]} & -0.4795 & 0.098 & -4.899 & 0.000 & -0.672 & -0.287 \\\\\n", "\\bottomrule\n", "\\end{tabular}\n", "%\\caption{OLS Regression Results}\n", "\\end{center}\n", "\n", "Notes: \\newline\n", " [1] Standard Errors assume that the covariance matrix of the errors is correctly specified." ], "text/plain": [ "\n", "\"\"\"\n", " OLS Regression Results \n", "==============================================================================\n", "Dep. Variable: rating R-squared: 0.039\n", "Model: OLS Adj. R-squared: 0.037\n", "No. Observations: 601 F-statistic: 24.00\n", "Covariance Type: nonrobust Prob (F-statistic): 1.24e-06\n", "===================================================================================\n", " coef std err t P>|t| [0.025 0.975]\n", "-----------------------------------------------------------------------------------\n", "Intercept 4.2749 0.083 51.635 0.000 4.112 4.437\n", "children[T.yes] -0.4795 0.098 -4.899 0.000 -0.672 -0.287\n", "===================================================================================\n", "\n", "Notes:\n", "[1] Standard Errors assume that the covariance matrix of the errors is correctly specified.\n", "\"\"\"" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "# First read in the data\n", "affairs = pd.read_csv('https://vincentarelbundock.github.io/Rdatasets/csv/AER/Affairs.csv')\n", "\n", "# Fit the simple model\n", "simple_model = smf.ols('rating ~ children', data=affairs).fit()\n", "\n", "# Simple summary\n", "display(simple_model.summary(slim=True))" ] }, { "cell_type": "markdown", "id": "0545b8a7-4a42-4c2d-9ef2-e5bdaf6a601a", "metadata": {}, "source": [ "### Using `marginaleffects` to get predictions\n", "With a fitted model, lets see some of the ways we can use `marginaleffects` to get some results.\n", "\n", "The first function is `me.predictions`, which is very general and returns a prediction for each datapoint that went into the data! Let's see how this works. `marginaleffects` returns a different type of dataframe (not a pandas one) so we just use the `to_pandas()` methods." ] }, { "cell_type": "code", "execution_count": 3, "id": "c84bd354-67df-48e3-8f02-bceb87a4cf0f", "metadata": {}, "outputs": [ { "data": { "text/html": [ "
\n", "\n", "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
rowidestimatestd_errorstatisticp_values_valueconf_lowconf_highrownamesaffairsgenderageyearsmarriedchildrenreligiousnesseducationoccupationrating
004.2748540.08279051.6347080.0inf4.1125884.43712040male37.010.0no31874
114.2748540.08279051.6347080.0inf4.1125884.43712050female27.04.0no41464
223.7953490.05220972.6956380.0inf3.6930223.897676110female32.015.0yes11214
\n", "
" ], "text/plain": [ " rowid estimate std_error statistic p_value s_value conf_low \\\n", "0 0 4.274854 0.082790 51.634708 0.0 inf 4.112588 \n", "1 1 4.274854 0.082790 51.634708 0.0 inf 4.112588 \n", "2 2 3.795349 0.052209 72.695638 0.0 inf 3.693022 \n", "\n", " conf_high rownames affairs gender age yearsmarried children \\\n", "0 4.437120 4 0 male 37.0 10.0 no \n", "1 4.437120 5 0 female 27.0 4.0 no \n", "2 3.897676 11 0 female 32.0 15.0 yes \n", "\n", " religiousness education occupation rating \n", "0 3 18 7 4 \n", "1 4 14 6 4 \n", "2 1 12 1 4 " ] }, "execution_count": 3, "metadata": {}, "output_type": "execute_result" } ], "source": [ "# Demonstrate prediction\n", "me.predictions(simple_model).to_pandas().head(3) # Predict for the fitted model" ] }, { "cell_type": "markdown", "id": "a74b1f4f-b0ae-4845-8a1d-4b6065536f69", "metadata": {}, "source": [ "Notice here we obtain, for each row/observation that went into the model:\n", "\n", "- an estimate (the prediction)\n", "- the std error\n", "- the t-statistic\n", "- the p-value (is it different from zero?)\n", "- s-value (we can ignore)\n", "- conf_low, conf_high - 95% confidence interval bounds\n", "\n", "... and the rest of the dataset so we can see what went in!" ] }, { "cell_type": "markdown", "id": "e87ae160-a6ee-4e01-a154-fd59d5a9a70d", "metadata": {}, "source": [ "This is great and can be very useful, but we want to use the model to *predict* an outcome for the values of its predictors. That is, we want to know, given someone has children, what is their satisfaction? What about if they do not have children? These are **conditional** predictions!\n", "\n", "To do this, we have to ask the model to give us its predictions, given its inputs. We need to generate our inputs, which we can do using the very powerful `me.datagrid` function. Lets explore this." ] }, { "cell_type": "markdown", "id": "cbbe43d1-cfdb-49a5-8fdf-62785041e131", "metadata": {}, "source": [ "### Generating inputs to get the right outputs\n", "`me.datagrid` allows us to specify the values of the predictors we want the models outputs for, in a simple way. We use the `me.datagrid` function, and the first argument is the name of the model, and the remaining arguments are the names of the predictors and the values we'd like them to take, specified in a list (and the values *must* be the same as what went into the model).\n", "\n", "So to generate a 'datagrid' (think - a set of questions we want to ask) for someone with and without children, we can do this:" ] }, { "cell_type": "code", "execution_count": 4, "id": "da3489ac-8a27-4c5d-8f91-7f6ae20f2b3f", "metadata": {}, "outputs": [ { "data": { "text/html": [ "
\n", "\n", "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
childrenrownamesaffairsgenderageyearsmarriedreligiousnesseducationoccupationrating
0yes8070female32.4875218.17769641455
1no8070female32.4875218.17769641455
\n", "
" ], "text/plain": [ " children rownames affairs gender age yearsmarried religiousness \\\n", "0 yes 807 0 female 32.487521 8.177696 4 \n", "1 no 807 0 female 32.487521 8.177696 4 \n", "\n", " education occupation rating \n", "0 14 5 5 \n", "1 14 5 5 " ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "# Demonstrate datagrid\n", "inputs = me.datagrid(simple_model,\n", " children=['yes', 'no'])\n", "\n", "display(inputs.to_pandas())" ] }, { "cell_type": "markdown", "id": "ac8e9c15-f357-4b22-a985-c1b1e6191dff", "metadata": {}, "source": [ "This basically generates data, based on the original data the model saw, that varies only in the predictor we are interested in - see the `children` column has 'yes', and 'no', rows, but is the same on everything else. We can then pass this onto the `me.predictions` function as part of the `newdata` argument, and obtain our conditional predictions!" ] }, { "cell_type": "code", "execution_count": 5, "id": "ce52f837-544e-402b-97e9-a686b0669386", "metadata": {}, "outputs": [ { "data": { "text/html": [ "
\n", "\n", "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
childrenrowidestimatestd_errorstatisticp_values_valueconf_lowconf_highrownamesaffairsgenderageyearsmarriedreligiousnesseducationoccupationrating
0yes03.7953490.05220972.6956380.0inf3.6930223.8976768070female32.4875218.17769641455
1no14.2748540.08279051.6347080.0inf4.1125884.4371208070female32.4875218.17769641455
\n", "
" ], "text/plain": [ " children rowid estimate std_error statistic p_value s_value conf_low \\\n", "0 yes 0 3.795349 0.052209 72.695638 0.0 inf 3.693022 \n", "1 no 1 4.274854 0.082790 51.634708 0.0 inf 4.112588 \n", "\n", " conf_high rownames affairs gender age yearsmarried \\\n", "0 3.897676 807 0 female 32.487521 8.177696 \n", "1 4.437120 807 0 female 32.487521 8.177696 \n", "\n", " religiousness education occupation rating \n", "0 4 14 5 5 \n", "1 4 14 5 5 " ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "# Obtain conditional predictions\n", "cond_pred = me.predictions(simple_model, newdata=inputs)\n", "display(cond_pred.to_pandas())" ] }, { "cell_type": "markdown", "id": "d121c86d-56ad-48d6-81a8-a5f72ae3e163", "metadata": {}, "source": [ "The `estimate` column reveals those with children have lower happiness than those without. " ] }, { "cell_type": "markdown", "id": "ad724ccf-5461-43ca-a6b8-b0d32d14fb83", "metadata": {}, "source": [ "### Comparing predictions\n", "We now see the mean satisfaction for marriage with children is 3.79, and without, it is 4.27. Now we may wish to know whether that difference is interesting. We can compare those values with a hypothesis test to work out the difference between them, and whether it is significant or not. This is achieved using the `hypothesis` argument, which takes varied inputs. We will see some complex ones later, but if we set this to the string `'pairwise'`, we will obtain a test of the difference between those predictions, as it will compare each row to every other row in the output!" ] }, { "cell_type": "code", "execution_count": 6, "id": "bda8f04e-18d4-4470-9f26-a02527aca426", "metadata": {}, "outputs": [ { "data": { "text/html": [ "
\n", "shape: (1, 8)
termestimatestd_errorstatisticp_values_valueconf_lowconf_high
strf64f64f64f64f64f64f64
"Row 1 - Row 2"-0.4795050.097877-4.8990359.6308e-719.985837-0.671341-0.287669
" ], "text/plain": [ "shape: (1, 8)\n", "┌───────────────┬──────────┬───────────┬──────┬──────────┬─────┬────────┬────────┐\n", "│ Term ┆ Estimate ┆ Std.Error ┆ z ┆ P(>|z|) ┆ S ┆ 2.5% ┆ 97.5% │\n", "│ --- ┆ --- ┆ --- ┆ --- ┆ --- ┆ --- ┆ --- ┆ --- │\n", "│ str ┆ str ┆ str ┆ str ┆ str ┆ str ┆ str ┆ str │\n", "╞═══════════════╪══════════╪═══════════╪══════╪══════════╪═════╪════════╪════════╡\n", "│ Row 1 - Row 2 ┆ -0.48 ┆ 0.0979 ┆ -4.9 ┆ 9.63e-07 ┆ 20 ┆ -0.671 ┆ -0.288 │\n", "└───────────────┴──────────┴───────────┴──────┴──────────┴─────┴────────┴────────┘\n", "\n", "Columns: term, estimate, std_error, statistic, p_value, s_value, conf_low, conf_high" ] }, "execution_count": 6, "metadata": {}, "output_type": "execute_result" } ], "source": [ "# Difference between our predictions\n", "me.predictions(simple_model, \n", " newdata=inputs, \n", " hypothesis='pairwise')" ] }, { "cell_type": "markdown", "id": "5a39fcdd-5d12-460c-8815-827ebbb92e7e", "metadata": {}, "source": [ "Amazingly and beautifully, this estimate is equal to the coefficient in the model! This is our test of interest, and we derived it entirely from a model - are those with children more satisfied in their marrige? No." ] }, { "cell_type": "markdown", "id": "df14eb06-1f69-4a85-9b29-5053c642d09b", "metadata": {}, "source": [ "### More complex models, the same approach\n", "While traditional approaches would require us to use our flow-chart to figure out which test to use, model-based inference makes tackling more complex problems easy. \n", "\n", "Lets extend the model to include the `yearsmarried` predictor. We will test whether the difference between those with and without children is present even when we adjust for the length of time they are married." ] }, { "cell_type": "code", "execution_count": 7, "id": "086da90e-cf1f-446a-8116-903f56e87a3e", "metadata": {}, "outputs": [ { "data": { "text/html": [ "\n", "\n", "\n", " \n", "\n", "\n", " \n", "\n", "\n", " \n", "\n", "\n", " \n", "\n", "
OLS Regression Results
Dep. Variable: rating R-squared: 0.064
Model: OLS Adj. R-squared: 0.061
No. Observations: 601 F-statistic: 20.43
Covariance Type: nonrobust Prob (F-statistic): 2.63e-09
\n", "\n", "\n", " \n", "\n", "\n", " \n", "\n", "\n", " \n", "\n", "\n", " \n", "\n", "
coef std err t P>|t| [0.025 0.975]
Intercept 4.0801 0.095 42.960 0.000 3.894 4.267
children[T.yes] -0.2073 0.118 -1.758 0.079 -0.439 0.024
scale(yearsmarried) -0.2144 0.053 -4.030 0.000 -0.319 -0.110


Notes:
[1] Standard Errors assume that the covariance matrix of the errors is correctly specified." ], "text/latex": [ "\\begin{center}\n", "\\begin{tabular}{lclc}\n", "\\toprule\n", "\\textbf{Dep. Variable:} & rating & \\textbf{ R-squared: } & 0.064 \\\\\n", "\\textbf{Model:} & OLS & \\textbf{ Adj. R-squared: } & 0.061 \\\\\n", "\\textbf{No. Observations:} & 601 & \\textbf{ F-statistic: } & 20.43 \\\\\n", "\\textbf{Covariance Type:} & nonrobust & \\textbf{ Prob (F-statistic):} & 2.63e-09 \\\\\n", "\\bottomrule\n", "\\end{tabular}\n", "\\begin{tabular}{lcccccc}\n", " & \\textbf{coef} & \\textbf{std err} & \\textbf{t} & \\textbf{P$> |$t$|$} & \\textbf{[0.025} & \\textbf{0.975]} \\\\\n", "\\midrule\n", "\\textbf{Intercept} & 4.0801 & 0.095 & 42.960 & 0.000 & 3.894 & 4.267 \\\\\n", "\\textbf{children[T.yes]} & -0.2073 & 0.118 & -1.758 & 0.079 & -0.439 & 0.024 \\\\\n", "\\textbf{scale(yearsmarried)} & -0.2144 & 0.053 & -4.030 & 0.000 & -0.319 & -0.110 \\\\\n", "\\bottomrule\n", "\\end{tabular}\n", "%\\caption{OLS Regression Results}\n", "\\end{center}\n", "\n", "Notes: \\newline\n", " [1] Standard Errors assume that the covariance matrix of the errors is correctly specified." ], "text/plain": [ "\n", "\"\"\"\n", " OLS Regression Results \n", "==============================================================================\n", "Dep. Variable: rating R-squared: 0.064\n", "Model: OLS Adj. R-squared: 0.061\n", "No. Observations: 601 F-statistic: 20.43\n", "Covariance Type: nonrobust Prob (F-statistic): 2.63e-09\n", "=======================================================================================\n", " coef std err t P>|t| [0.025 0.975]\n", "---------------------------------------------------------------------------------------\n", "Intercept 4.0801 0.095 42.960 0.000 3.894 4.267\n", "children[T.yes] -0.2073 0.118 -1.758 0.079 -0.439 0.024\n", "scale(yearsmarried) -0.2144 0.053 -4.030 0.000 -0.319 -0.110\n", "=======================================================================================\n", "\n", "Notes:\n", "[1] Standard Errors assume that the covariance matrix of the errors is correctly specified.\n", "\"\"\"" ] }, "execution_count": 7, "metadata": {}, "output_type": "execute_result" } ], "source": [ "# Build a new model\n", "model2 = smf.ols('rating ~ children + scale(yearsmarried)', data=affairs).fit() # note 'scale'\n", "model2.summary(slim=True)" ] }, { "cell_type": "markdown", "id": "67b276bd-cf5c-43f4-8f11-015ba2ed67f8", "metadata": {}, "source": [ "If we want to make *marginal predictions*, that is a prediction while we hold other variables constant, we simply *leave out* the variables we want to hold constant. We repeat the same procedure as before:" ] }, { "cell_type": "code", "execution_count": 8, "id": "aacbb967-ae14-432e-8707-ea67115cf2ae", "metadata": {}, "outputs": [ { "data": { "text/plain": [ "'Predict this'" ] }, "metadata": {}, "output_type": "display_data" }, { "data": { "text/html": [ "
\n", "shape: (2, 10)
childrenyearsmarriedrownamesaffairsgenderagereligiousnesseducationoccupationrating
stri64i64i64strf64i64i64i64i64
"yes"1018430"female"32.48752141455
"no"1018430"female"32.48752141455
" ], "text/plain": [ "shape: (2, 10)\n", "┌──────────┬──────────────┬──────────┬─────────┬───┬─────────────┬───────────┬────────────┬────────┐\n", "│ children ┆ yearsmarried ┆ rownames ┆ affairs ┆ … ┆ religiousne ┆ education ┆ occupation ┆ rating │\n", "│ --- ┆ --- ┆ --- ┆ --- ┆ ┆ ss ┆ --- ┆ --- ┆ --- │\n", "│ str ┆ i64 ┆ i64 ┆ i64 ┆ ┆ --- ┆ i64 ┆ i64 ┆ i64 │\n", "│ ┆ ┆ ┆ ┆ ┆ i64 ┆ ┆ ┆ │\n", "╞══════════╪══════════════╪══════════╪═════════╪═══╪═════════════╪═══════════╪════════════╪════════╡\n", "│ yes ┆ 10 ┆ 1843 ┆ 0 ┆ … ┆ 4 ┆ 14 ┆ 5 ┆ 5 │\n", "│ no ┆ 10 ┆ 1843 ┆ 0 ┆ … ┆ 4 ┆ 14 ┆ 5 ┆ 5 │\n", "└──────────┴──────────────┴──────────┴─────────┴───┴─────────────┴───────────┴────────────┴────────┘" ] }, "metadata": {}, "output_type": "display_data" }, { "data": { "text/plain": [ "'Predictions'" ] }, "metadata": {}, "output_type": "display_data" }, { "data": { "text/html": [ "
\n", "shape: (2, 18)
childrenyearsmarriedrowidestimatestd_errorstatisticp_values_valueconf_lowconf_highrownamesaffairsgenderagereligiousnesseducationoccupationrating
stri64i32f64f64f64f64f64f64f64i64i64strf64i64i64i64i64
"yes"1003.8026150.05158973.7104270.0inf3.7015043.90372718430"female"32.48752141455
"no"1014.0098980.10491538.2204160.0inf3.8042694.21552818430"female"32.48752141455
" ], "text/plain": [ "shape: (2, 9)\n", "┌──────────┬──────────────┬──────────┬───────────┬───┬─────────┬─────┬──────┬───────┐\n", "│ children ┆ yearsmarried ┆ Estimate ┆ Std.Error ┆ … ┆ P(>|z|) ┆ S ┆ 2.5% ┆ 97.5% │\n", "│ --- ┆ --- ┆ --- ┆ --- ┆ ┆ --- ┆ --- ┆ --- ┆ --- │\n", "│ str ┆ str ┆ str ┆ str ┆ ┆ str ┆ str ┆ str ┆ str │\n", "╞══════════╪══════════════╪══════════╪═══════════╪═══╪═════════╪═════╪══════╪═══════╡\n", "│ yes ┆ 10 ┆ 3.8 ┆ 0.0516 ┆ … ┆ 0 ┆ inf ┆ 3.7 ┆ 3.9 │\n", "│ no ┆ 10 ┆ 4.01 ┆ 0.105 ┆ … ┆ 0 ┆ inf ┆ 3.8 ┆ 4.22 │\n", "└──────────┴──────────────┴──────────┴───────────┴───┴─────────┴─────┴──────┴───────┘\n", "\n", "Columns: children, yearsmarried, rowid, estimate, std_error, statistic, p_value, s_value, conf_low, conf_high, rownames, affairs, gender, age, religiousness, education, occupation, rating" ] }, "metadata": {}, "output_type": "display_data" }, { "data": { "text/plain": [ "'The difference:'" ] }, "metadata": {}, "output_type": "display_data" }, { "data": { "text/html": [ "
\n", "shape: (1, 8)
termestimatestd_errorstatisticp_values_valueconf_lowconf_high
strf64f64f64f64f64f64f64
"Row 1 - Row 2"-0.2072830.117922-1.7577930.0787833.665976-0.4384070.02384
" ], "text/plain": [ "shape: (1, 8)\n", "┌───────────────┬──────────┬───────────┬───────┬─────────┬──────┬────────┬────────┐\n", "│ Term ┆ Estimate ┆ Std.Error ┆ z ┆ P(>|z|) ┆ S ┆ 2.5% ┆ 97.5% │\n", "│ --- ┆ --- ┆ --- ┆ --- ┆ --- ┆ --- ┆ --- ┆ --- │\n", "│ str ┆ str ┆ str ┆ str ┆ str ┆ str ┆ str ┆ str │\n", "╞═══════════════╪══════════╪═══════════╪═══════╪═════════╪══════╪════════╪════════╡\n", "│ Row 1 - Row 2 ┆ -0.207 ┆ 0.118 ┆ -1.76 ┆ 0.0788 ┆ 3.67 ┆ -0.438 ┆ 0.0238 │\n", "└───────────────┴──────────┴───────────┴───────┴─────────┴──────┴────────┴────────┘\n", "\n", "Columns: term, estimate, std_error, statistic, p_value, s_value, conf_low, conf_high" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "# Make a datagrid of what we want to know\n", "new_datagrid = me.datagrid(model2, children=['yes', 'no'], yearsmarried=[10]) # Notice NO yearsmarried variable\n", "display('Predict this', new_datagrid)\n", "\n", "# Model makes predictions on this\n", "new_predictions = me.predictions(model2, newdata=new_datagrid)\n", "display('Predictions', new_predictions)\n", "\n", "# The contrast is done as before\n", "contrast = me.predictions(model2, newdata=new_datagrid, hypothesis='pairwise')\n", "display('The difference:', contrast)" ] }, { "cell_type": "markdown", "id": "143cba94-17df-405e-ab07-4e8caf551b73", "metadata": {}, "source": [ "### Exploring interactions with greater complexity\n", "A huge bonus of this approach is that it makes answering complex questions seamless. Lets build a model that includes an interaction between children and gender, and includes the `yearsmarried` covariate. That is - whats the difference in marital satisfaction between men and women who do and dont have children, controlling for how long they have been married? \n", "\n", "This will introduce us to some subtle ways of working with the `me.predictions` and particularly the hypothesis approach." ] }, { "cell_type": "code", "execution_count": 9, "id": "7d448893-4e03-4dd2-9a99-1100f5b750a2", "metadata": {}, "outputs": [ { "data": { "text/html": [ "\n", "\n", "\n", " \n", "\n", "\n", " \n", "\n", "\n", " \n", "\n", "\n", " \n", "\n", "
OLS Regression Results
Dep. Variable: rating R-squared: 0.070
Model: OLS Adj. R-squared: 0.063
No. Observations: 601 F-statistic: 11.16
Covariance Type: nonrobust Prob (F-statistic): 9.72e-09
\n", "\n", "\n", " \n", "\n", "\n", " \n", "\n", "\n", " \n", "\n", "\n", " \n", "\n", "\n", " \n", "\n", "\n", " \n", "\n", "
coef std err t P>|t| [0.025 0.975]
Intercept 4.1985 0.120 35.013 0.000 3.963 4.434
gender[T.male] -0.2607 0.166 -1.572 0.116 -0.586 0.065
children[T.yes] -0.3860 0.150 -2.571 0.010 -0.681 -0.091
gender[T.male]:children[T.yes] 0.3751 0.196 1.917 0.056 -0.009 0.759
scale(yearsmarried) -0.2049 0.053 -3.840 0.000 -0.310 -0.100


Notes:
[1] Standard Errors assume that the covariance matrix of the errors is correctly specified." ], "text/latex": [ "\\begin{center}\n", "\\begin{tabular}{lclc}\n", "\\toprule\n", "\\textbf{Dep. Variable:} & rating & \\textbf{ R-squared: } & 0.070 \\\\\n", "\\textbf{Model:} & OLS & \\textbf{ Adj. R-squared: } & 0.063 \\\\\n", "\\textbf{No. Observations:} & 601 & \\textbf{ F-statistic: } & 11.16 \\\\\n", "\\textbf{Covariance Type:} & nonrobust & \\textbf{ Prob (F-statistic):} & 9.72e-09 \\\\\n", "\\bottomrule\n", "\\end{tabular}\n", "\\begin{tabular}{lcccccc}\n", " & \\textbf{coef} & \\textbf{std err} & \\textbf{t} & \\textbf{P$> |$t$|$} & \\textbf{[0.025} & \\textbf{0.975]} \\\\\n", "\\midrule\n", "\\textbf{Intercept} & 4.1985 & 0.120 & 35.013 & 0.000 & 3.963 & 4.434 \\\\\n", "\\textbf{gender[T.male]} & -0.2607 & 0.166 & -1.572 & 0.116 & -0.586 & 0.065 \\\\\n", "\\textbf{children[T.yes]} & -0.3860 & 0.150 & -2.571 & 0.010 & -0.681 & -0.091 \\\\\n", "\\textbf{gender[T.male]:children[T.yes]} & 0.3751 & 0.196 & 1.917 & 0.056 & -0.009 & 0.759 \\\\\n", "\\textbf{scale(yearsmarried)} & -0.2049 & 0.053 & -3.840 & 0.000 & -0.310 & -0.100 \\\\\n", "\\bottomrule\n", "\\end{tabular}\n", "%\\caption{OLS Regression Results}\n", "\\end{center}\n", "\n", "Notes: \\newline\n", " [1] Standard Errors assume that the covariance matrix of the errors is correctly specified." ], "text/plain": [ "\n", "\"\"\"\n", " OLS Regression Results \n", "==============================================================================\n", "Dep. Variable: rating R-squared: 0.070\n", "Model: OLS Adj. R-squared: 0.063\n", "No. Observations: 601 F-statistic: 11.16\n", "Covariance Type: nonrobust Prob (F-statistic): 9.72e-09\n", "==================================================================================================\n", " coef std err t P>|t| [0.025 0.975]\n", "--------------------------------------------------------------------------------------------------\n", "Intercept 4.1985 0.120 35.013 0.000 3.963 4.434\n", "gender[T.male] -0.2607 0.166 -1.572 0.116 -0.586 0.065\n", "children[T.yes] -0.3860 0.150 -2.571 0.010 -0.681 -0.091\n", "gender[T.male]:children[T.yes] 0.3751 0.196 1.917 0.056 -0.009 0.759\n", "scale(yearsmarried) -0.2049 0.053 -3.840 0.000 -0.310 -0.100\n", "==================================================================================================\n", "\n", "Notes:\n", "[1] Standard Errors assume that the covariance matrix of the errors is correctly specified.\n", "\"\"\"" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "# Specify the model\n", "model3 = smf.ols('rating ~ gender * children + scale(yearsmarried)', data=affairs).fit()\n", "display(model3.summary(slim=True))" ] }, { "cell_type": "code", "execution_count": 10, "id": "c69d3dd0-b02b-4cdd-b804-6a3469e3bf6b", "metadata": {}, "outputs": [ { "data": { "text/html": [ "
\n", "shape: (4, 10)
childrengenderrownamesaffairsageyearsmarriedreligiousnesseducationoccupationrating
strstri64i64f64f64i64i64i64i64
"yes""male"1304032.4875218.17769641455
"yes""female"1304032.4875218.17769641455
"no""male"1304032.4875218.17769641455
"no""female"1304032.4875218.17769641455
" ], "text/plain": [ "shape: (4, 10)\n", "┌──────────┬────────┬──────────┬─────────┬───┬───────────────┬───────────┬────────────┬────────┐\n", "│ children ┆ gender ┆ rownames ┆ affairs ┆ … ┆ religiousness ┆ education ┆ occupation ┆ rating │\n", "│ --- ┆ --- ┆ --- ┆ --- ┆ ┆ --- ┆ --- ┆ --- ┆ --- │\n", "│ str ┆ str ┆ i64 ┆ i64 ┆ ┆ i64 ┆ i64 ┆ i64 ┆ i64 │\n", "╞══════════╪════════╪══════════╪═════════╪═══╪═══════════════╪═══════════╪════════════╪════════╡\n", "│ yes ┆ male ┆ 1304 ┆ 0 ┆ … ┆ 4 ┆ 14 ┆ 5 ┆ 5 │\n", "│ yes ┆ female ┆ 1304 ┆ 0 ┆ … ┆ 4 ┆ 14 ┆ 5 ┆ 5 │\n", "│ no ┆ male ┆ 1304 ┆ 0 ┆ … ┆ 4 ┆ 14 ┆ 5 ┆ 5 │\n", "│ no ┆ female ┆ 1304 ┆ 0 ┆ … ┆ 4 ┆ 14 ┆ 5 ┆ 5 │\n", "└──────────┴────────┴──────────┴─────────┴───┴───────────────┴───────────┴────────────┴────────┘" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "# Build a datagrid to answer our question\n", "datagrid3 = me.datagrid(model3, \n", " children=['yes', 'no'], \n", " gender=['male', 'female'])\n", "display(datagrid3)" ] }, { "cell_type": "code", "execution_count": 11, "id": "b1b9543d-359d-4cab-952e-28f3d74063c0", "metadata": {}, "outputs": [ { "data": { "text/html": [ "
\n", "shape: (4, 18)
childrengenderrowidestimatestd_errorstatisticp_values_valueconf_lowconf_highrownamesaffairsageyearsmarriedreligiousnesseducationoccupationrating
strstri32f64f64f64f64f64f64f64i64i64f64f64i64i64i64i64
"yes""male"03.9268210.07476452.5229260.0inf3.7802874.0733561304032.4875218.17769641455
"yes""female"13.8124410.07598550.1738710.0inf3.6635143.9613681304032.4875218.17769641455
"no""male"23.9378130.13249129.7212720.0inf3.6781344.1974911304032.4875218.17769641455
"no""female"34.198490.11991235.0131040.0inf3.9634674.4335131304032.4875218.17769641455
" ], "text/plain": [ "shape: (4, 9)\n", "┌──────────┬────────┬──────────┬───────────┬───┬─────────┬─────┬──────┬───────┐\n", "│ children ┆ gender ┆ Estimate ┆ Std.Error ┆ … ┆ P(>|z|) ┆ S ┆ 2.5% ┆ 97.5% │\n", "│ --- ┆ --- ┆ --- ┆ --- ┆ ┆ --- ┆ --- ┆ --- ┆ --- │\n", "│ str ┆ str ┆ str ┆ str ┆ ┆ str ┆ str ┆ str ┆ str │\n", "╞══════════╪════════╪══════════╪═══════════╪═══╪═════════╪═════╪══════╪═══════╡\n", "│ yes ┆ male ┆ 3.93 ┆ 0.0748 ┆ … ┆ 0 ┆ inf ┆ 3.78 ┆ 4.07 │\n", "│ yes ┆ female ┆ 3.81 ┆ 0.076 ┆ … ┆ 0 ┆ inf ┆ 3.66 ┆ 3.96 │\n", "│ no ┆ male ┆ 3.94 ┆ 0.132 ┆ … ┆ 0 ┆ inf ┆ 3.68 ┆ 4.2 │\n", "│ no ┆ female ┆ 4.2 ┆ 0.12 ┆ … ┆ 0 ┆ inf ┆ 3.96 ┆ 4.43 │\n", "└──────────┴────────┴──────────┴───────────┴───┴─────────┴─────┴──────┴───────┘\n", "\n", "Columns: children, gender, rowid, estimate, std_error, statistic, p_value, s_value, conf_low, conf_high, rownames, affairs, age, yearsmarried, religiousness, education, occupation, rating" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "# Obtain predictions\n", "predictions3 = me.predictions(model3, newdata=datagrid3)\n", "display(predictions3)" ] }, { "cell_type": "markdown", "id": "c87cd937-1451-44cc-abf3-f111355de3d3", "metadata": {}, "source": [ "We can make a plot of these if we like with `seaborn`." ] }, { "cell_type": "code", "execution_count": 12, "id": "b6514a3c-2e03-4c9e-a466-6a16fb28a501", "metadata": {}, "outputs": [ { "data": { "text/plain": [ "" ] }, "execution_count": 12, "metadata": {}, "output_type": "execute_result" }, { "data": { "image/png": "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", "text/plain": [ "
" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "# visualise\n", "sns.lineplot(data=predictions3,\n", " x='children', y='estimate',\n", " hue='gender')" ] }, { "cell_type": "markdown", "id": "43ee1dbe-9715-4511-b1ad-0f65746e270a", "metadata": {}, "source": [ "The output of our predictions is essentially four rows, one for each combination of gender and child status (e.g. female with children, male without children, and so on). \n", "\n", "We can ask for specific differences here by indicating the *row labels* to compare and ask whether they are equal. The row labels are prefaced with the letter 'b'. This is a little confusing, so lets see how this works. Here are the predictions again:" ] }, { "cell_type": "code", "execution_count": 13, "id": "857cb407-2a35-4500-be2e-f17dac6d544c", "metadata": {}, "outputs": [ { "data": { "text/html": [ "
\n", "shape: (4, 18)
childrengenderrowidestimatestd_errorstatisticp_values_valueconf_lowconf_highrownamesaffairsageyearsmarriedreligiousnesseducationoccupationrating
strstri32f64f64f64f64f64f64f64i64i64f64f64i64i64i64i64
"yes""male"03.9268210.07476452.5229260.0inf3.7802874.0733561304032.4875218.17769641455
"yes""female"13.8124410.07598550.1738710.0inf3.6635143.9613681304032.4875218.17769641455
"no""male"23.9378130.13249129.7212720.0inf3.6781344.1974911304032.4875218.17769641455
"no""female"34.198490.11991235.0131040.0inf3.9634674.4335131304032.4875218.17769641455
" ], "text/plain": [ "shape: (4, 9)\n", "┌──────────┬────────┬──────────┬───────────┬───┬─────────┬─────┬──────┬───────┐\n", "│ children ┆ gender ┆ Estimate ┆ Std.Error ┆ … ┆ P(>|z|) ┆ S ┆ 2.5% ┆ 97.5% │\n", "│ --- ┆ --- ┆ --- ┆ --- ┆ ┆ --- ┆ --- ┆ --- ┆ --- │\n", "│ str ┆ str ┆ str ┆ str ┆ ┆ str ┆ str ┆ str ┆ str │\n", "╞══════════╪════════╪══════════╪═══════════╪═══╪═════════╪═════╪══════╪═══════╡\n", "│ yes ┆ male ┆ 3.93 ┆ 0.0748 ┆ … ┆ 0 ┆ inf ┆ 3.78 ┆ 4.07 │\n", "│ yes ┆ female ┆ 3.81 ┆ 0.076 ┆ … ┆ 0 ┆ inf ┆ 3.66 ┆ 3.96 │\n", "│ no ┆ male ┆ 3.94 ┆ 0.132 ┆ … ┆ 0 ┆ inf ┆ 3.68 ┆ 4.2 │\n", "│ no ┆ female ┆ 4.2 ┆ 0.12 ┆ … ┆ 0 ┆ inf ┆ 3.96 ┆ 4.43 │\n", "└──────────┴────────┴──────────┴───────────┴───┴─────────┴─────┴──────┴───────┘\n", "\n", "Columns: children, gender, rowid, estimate, std_error, statistic, p_value, s_value, conf_low, conf_high, rownames, affairs, age, yearsmarried, religiousness, education, occupation, rating" ] }, "execution_count": 13, "metadata": {}, "output_type": "execute_result" } ], "source": [ "# Predictions\n", "me.predictions(model3, \n", " newdata=datagrid3)" ] }, { "cell_type": "markdown", "id": "16092629-48bf-4b4c-92f8-5b2106852cec", "metadata": {}, "source": [ "If I want to compare males with children (ROW 1) to males without children (ROW 3), then this is the way:" ] }, { "cell_type": "code", "execution_count": 14, "id": "37ef143e-b555-4e3a-ac3f-4608acfdb027", "metadata": {}, "outputs": [ { "data": { "text/html": [ "
\n", "shape: (1, 8)
termestimatestd_errorstatisticp_values_valueconf_lowconf_high
strf64f64f64f64f64f64f64
"b1=b3"-0.0109910.156498-0.0702330.9440080.083129-0.3177220.295739
" ], "text/plain": [ "shape: (1, 8)\n", "┌───────┬──────────┬───────────┬─────────┬─────────┬────────┬────────┬───────┐\n", "│ Term ┆ Estimate ┆ Std.Error ┆ z ┆ P(>|z|) ┆ S ┆ 2.5% ┆ 97.5% │\n", "│ --- ┆ --- ┆ --- ┆ --- ┆ --- ┆ --- ┆ --- ┆ --- │\n", "│ str ┆ str ┆ str ┆ str ┆ str ┆ str ┆ str ┆ str │\n", "╞═══════╪══════════╪═══════════╪═════════╪═════════╪════════╪════════╪═══════╡\n", "│ b1=b3 ┆ -0.011 ┆ 0.156 ┆ -0.0702 ┆ 0.944 ┆ 0.0831 ┆ -0.318 ┆ 0.296 │\n", "└───────┴──────────┴───────────┴─────────┴─────────┴────────┴────────┴───────┘\n", "\n", "Columns: term, estimate, std_error, statistic, p_value, s_value, conf_low, conf_high" ] }, "execution_count": 14, "metadata": {}, "output_type": "execute_result" } ], "source": [ "# A specific difference\n", "me.predictions(model3, \n", " newdata=datagrid3,\n", " hypothesis='b1 = b3')" ] }, { "cell_type": "markdown", "id": "e8765443-6c1d-44dc-ab79-61e3db86a9a0", "metadata": {}, "source": [ "Are there differences between females with and without children? That's row 2 and row 4, so:" ] }, { "cell_type": "code", "execution_count": 15, "id": "ec4e153a-ca2e-4738-bda2-1a2288f42e4e", "metadata": {}, "outputs": [ { "data": { "text/html": [ "
\n", "shape: (1, 8)
termestimatestd_errorstatisticp_values_valueconf_lowconf_high
strf64f64f64f64f64f64f64
"b2=b4"-0.3860490.150131-2.5714190.0101286.625468-0.6803-0.091798
" ], "text/plain": [ "shape: (1, 8)\n", "┌───────┬──────────┬───────────┬───────┬─────────┬──────┬───────┬─────────┐\n", "│ Term ┆ Estimate ┆ Std.Error ┆ z ┆ P(>|z|) ┆ S ┆ 2.5% ┆ 97.5% │\n", "│ --- ┆ --- ┆ --- ┆ --- ┆ --- ┆ --- ┆ --- ┆ --- │\n", "│ str ┆ str ┆ str ┆ str ┆ str ┆ str ┆ str ┆ str │\n", "╞═══════╪══════════╪═══════════╪═══════╪═════════╪══════╪═══════╪═════════╡\n", "│ b2=b4 ┆ -0.386 ┆ 0.15 ┆ -2.57 ┆ 0.0101 ┆ 6.63 ┆ -0.68 ┆ -0.0918 │\n", "└───────┴──────────┴───────────┴───────┴─────────┴──────┴───────┴─────────┘\n", "\n", "Columns: term, estimate, std_error, statistic, p_value, s_value, conf_low, conf_high" ] }, "execution_count": 15, "metadata": {}, "output_type": "execute_result" } ], "source": [ "# Another\n", "me.predictions(model3, \n", " newdata=datagrid3,\n", " hypothesis='b2 = b4')" ] }, { "cell_type": "markdown", "id": "bc7a83eb-9bc5-4fbb-8a26-2cad82de5ef1", "metadata": {}, "source": [ "You might also be interested in the difference between men and women, averaged over whether they have children or not. You can ask `marginaleffects` to do this for you by including the variable you wish to target in the 'by' keyword. For example, if I want to see the predictions for men and women averaged across their child status, this is the way:" ] }, { "cell_type": "code", "execution_count": 16, "id": "aa097e21-67aa-4bb1-bdb7-3d775645b7ee", "metadata": {}, "outputs": [ { "data": { "text/html": [ "
\n", "shape: (2, 8)
genderestimatestd_errorstatisticp_values_valueconf_lowconf_high
strf64f64f64f64f64f64f64
"female"4.0054650.06664460.1021610.0inf3.8748454.136086
"male"3.9323170.07381753.2714530.0inf3.7876394.076995
" ], "text/plain": [ "shape: (2, 8)\n", "┌────────┬──────────┬───────────┬──────┬─────────┬─────┬──────┬───────┐\n", "│ gender ┆ Estimate ┆ Std.Error ┆ z ┆ P(>|z|) ┆ S ┆ 2.5% ┆ 97.5% │\n", "│ --- ┆ --- ┆ --- ┆ --- ┆ --- ┆ --- ┆ --- ┆ --- │\n", "│ str ┆ str ┆ str ┆ str ┆ str ┆ str ┆ str ┆ str │\n", "╞════════╪══════════╪═══════════╪══════╪═════════╪═════╪══════╪═══════╡\n", "│ female ┆ 4.01 ┆ 0.0666 ┆ 60.1 ┆ 0 ┆ inf ┆ 3.87 ┆ 4.14 │\n", "│ male ┆ 3.93 ┆ 0.0738 ┆ 53.3 ┆ 0 ┆ inf ┆ 3.79 ┆ 4.08 │\n", "└────────┴──────────┴───────────┴──────┴─────────┴─────┴──────┴───────┘\n", "\n", "Columns: gender, estimate, std_error, statistic, p_value, s_value, conf_low, conf_high" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "# Averaging over something\n", "avg_over = me.predictions(model3,\n", " newdata=datagrid3,\n", " by='gender')\n", "display(avg_over)" ] }, { "cell_type": "markdown", "id": "b1189642-c031-495b-ae10-819313197958", "metadata": {}, "source": [ "Testing differences between *those* predictions proceeds as normal:" ] }, { "cell_type": "code", "execution_count": 17, "id": "477b5e1d-dbb8-4403-911d-97661c87d890", "metadata": {}, "outputs": [ { "data": { "text/html": [ "
\n", "shape: (1, 8)
termestimatestd_errorstatisticp_values_valueconf_lowconf_high
strf64f64f64f64f64f64f64
"b1=b2"0.0731480.0974440.7506670.4528531.142885-0.1178390.264136
" ], "text/plain": [ "shape: (1, 8)\n", "┌───────┬──────────┬───────────┬───────┬─────────┬──────┬────────┬───────┐\n", "│ Term ┆ Estimate ┆ Std.Error ┆ z ┆ P(>|z|) ┆ S ┆ 2.5% ┆ 97.5% │\n", "│ --- ┆ --- ┆ --- ┆ --- ┆ --- ┆ --- ┆ --- ┆ --- │\n", "│ str ┆ str ┆ str ┆ str ┆ str ┆ str ┆ str ┆ str │\n", "╞═══════╪══════════╪═══════════╪═══════╪═════════╪══════╪════════╪═══════╡\n", "│ b1=b2 ┆ 0.0731 ┆ 0.0974 ┆ 0.751 ┆ 0.453 ┆ 1.14 ┆ -0.118 ┆ 0.264 │\n", "└───────┴──────────┴───────────┴───────┴─────────┴──────┴────────┴───────┘\n", "\n", "Columns: term, estimate, std_error, statistic, p_value, s_value, conf_low, conf_high" ] }, "execution_count": 17, "metadata": {}, "output_type": "execute_result" } ], "source": [ "# Differences\n", "me.predictions(model3,\n", " newdata=datagrid3,\n", " by='gender',\n", " hypothesis='b1=b2') # 'pairwise' also works here" ] }, { "cell_type": "markdown", "id": "4002f71d-5648-49f8-a556-964cb5f78183", "metadata": {}, "source": [ "### Exploring more interactions - categorical and continuous\n", "What if we have categorical and continuous predictors interacting, and want to use our marginal effects to understand what is going on?\n", "\n", "When we are working with continuous variables, we have two options - the comparison approach as we did above, or the simple slopes approach. Lets first explore the comparison approach." ] }, { "cell_type": "markdown", "id": "dc125c1c-aa59-4a14-9258-f400e2d8cc16", "metadata": {}, "source": [ "#### Exploring more interactions - categorical and continuous - **comparisons** \n", "When working with continuous variables and predictions, we need to provide some candidate values across the range of the predictor for the model to work with. Lets see how this works by fitting a model that predicts marital satisfaction from the interaction of gender and years-married. That is, how does satisfaction change for females and males over the course of a marriage?\n", "\n", "First we fit the model:" ] }, { "cell_type": "code", "execution_count": 18, "id": "8f6a3206-5409-447f-b450-ee6c7ee87f5e", "metadata": {}, "outputs": [ { "data": { "text/html": [ "\n", "\n", "\n", " \n", "\n", "\n", " \n", "\n", "\n", " \n", "\n", "\n", " \n", "\n", "
OLS Regression Results
Dep. Variable: rating R-squared: 0.066
Model: OLS Adj. R-squared: 0.061
No. Observations: 601 F-statistic: 14.02
Covariance Type: nonrobust Prob (F-statistic): 7.68e-09
\n", "\n", "\n", " \n", "\n", "\n", " \n", "\n", "\n", " \n", "\n", "\n", " \n", "\n", "\n", " \n", "\n", "
coef std err t P>|t| [0.025 0.975]
Intercept 4.4470 0.105 42.374 0.000 4.241 4.653
gender[T.male] -0.2668 0.156 -1.715 0.087 -0.572 0.039
yearsmarried -0.0633 0.011 -5.902 0.000 -0.084 -0.042
gender[T.male]:yearsmarried 0.0325 0.016 2.069 0.039 0.002 0.063


Notes:
[1] Standard Errors assume that the covariance matrix of the errors is correctly specified." ], "text/latex": [ "\\begin{center}\n", "\\begin{tabular}{lclc}\n", "\\toprule\n", "\\textbf{Dep. Variable:} & rating & \\textbf{ R-squared: } & 0.066 \\\\\n", "\\textbf{Model:} & OLS & \\textbf{ Adj. R-squared: } & 0.061 \\\\\n", "\\textbf{No. Observations:} & 601 & \\textbf{ F-statistic: } & 14.02 \\\\\n", "\\textbf{Covariance Type:} & nonrobust & \\textbf{ Prob (F-statistic):} & 7.68e-09 \\\\\n", "\\bottomrule\n", "\\end{tabular}\n", "\\begin{tabular}{lcccccc}\n", " & \\textbf{coef} & \\textbf{std err} & \\textbf{t} & \\textbf{P$> |$t$|$} & \\textbf{[0.025} & \\textbf{0.975]} \\\\\n", "\\midrule\n", "\\textbf{Intercept} & 4.4470 & 0.105 & 42.374 & 0.000 & 4.241 & 4.653 \\\\\n", "\\textbf{gender[T.male]} & -0.2668 & 0.156 & -1.715 & 0.087 & -0.572 & 0.039 \\\\\n", "\\textbf{yearsmarried} & -0.0633 & 0.011 & -5.902 & 0.000 & -0.084 & -0.042 \\\\\n", "\\textbf{gender[T.male]:yearsmarried} & 0.0325 & 0.016 & 2.069 & 0.039 & 0.002 & 0.063 \\\\\n", "\\bottomrule\n", "\\end{tabular}\n", "%\\caption{OLS Regression Results}\n", "\\end{center}\n", "\n", "Notes: \\newline\n", " [1] Standard Errors assume that the covariance matrix of the errors is correctly specified." ], "text/plain": [ "\n", "\"\"\"\n", " OLS Regression Results \n", "==============================================================================\n", "Dep. Variable: rating R-squared: 0.066\n", "Model: OLS Adj. R-squared: 0.061\n", "No. Observations: 601 F-statistic: 14.02\n", "Covariance Type: nonrobust Prob (F-statistic): 7.68e-09\n", "===============================================================================================\n", " coef std err t P>|t| [0.025 0.975]\n", "-----------------------------------------------------------------------------------------------\n", "Intercept 4.4470 0.105 42.374 0.000 4.241 4.653\n", "gender[T.male] -0.2668 0.156 -1.715 0.087 -0.572 0.039\n", "yearsmarried -0.0633 0.011 -5.902 0.000 -0.084 -0.042\n", "gender[T.male]:yearsmarried 0.0325 0.016 2.069 0.039 0.002 0.063\n", "===============================================================================================\n", "\n", "Notes:\n", "[1] Standard Errors assume that the covariance matrix of the errors is correctly specified.\n", "\"\"\"" ] }, "execution_count": 18, "metadata": {}, "output_type": "execute_result" } ], "source": [ "# New model\n", "model4 = smf.ols('rating ~ gender * yearsmarried', data=affairs).fit()\n", "model4.summary(slim=True)" ] }, { "cell_type": "markdown", "id": "21b05aa0-8f60-4dd5-9884-00dddf77def3", "metadata": {}, "source": [ "We need to provide some values for `yearsmarried` for our model to predict for. What should we choose? There's no right or wrong answer - we might have some specific ideas in mind (e.g., 6 months, 2 years, 10 years, and 20 years), or we could provide something like the interquartile range for the `yearsmarried` variable and use those. Pandas can calculate that for us easily. Here we collect the 25th, 50th, and 75th percentiles of years married:" ] }, { "cell_type": "code", "execution_count": 19, "id": "ad77ac25-bd97-4733-bd6c-31f28f61b185", "metadata": {}, "outputs": [ { "data": { "text/plain": [ "0.25 4.0\n", "0.50 7.0\n", "0.75 15.0\n", "Name: yearsmarried, dtype: float64" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "# Use pandas to get the quartiles\n", "quartiles = affairs['yearsmarried'].quantile([.25, .5, .75])\n", "display(quartiles)" ] }, { "cell_type": "markdown", "id": "995d8742-c5ad-4ff3-9106-9f22039fa861", "metadata": {}, "source": [ "And now we pass those back into the datagrid approach, also asking for predictions on males and females:" ] }, { "cell_type": "code", "execution_count": 20, "id": "c8c75789-1a15-4703-ab02-caabdbf62ec4", "metadata": {}, "outputs": [ { "data": { "text/html": [ "
\n", "shape: (6, 10)
genderyearsmarriedrownamesaffairsagechildrenreligiousnesseducationoccupationrating
strf64i64i64f64stri64i64i64i64
"male"4.0228032.487521"yes"41455
"male"7.0228032.487521"yes"41455
"male"15.0228032.487521"yes"41455
"female"4.0228032.487521"yes"41455
"female"7.0228032.487521"yes"41455
"female"15.0228032.487521"yes"41455
" ], "text/plain": [ "shape: (6, 10)\n", "┌────────┬──────────────┬──────────┬─────────┬───┬───────────────┬───────────┬────────────┬────────┐\n", "│ gender ┆ yearsmarried ┆ rownames ┆ affairs ┆ … ┆ religiousness ┆ education ┆ occupation ┆ rating │\n", "│ --- ┆ --- ┆ --- ┆ --- ┆ ┆ --- ┆ --- ┆ --- ┆ --- │\n", "│ str ┆ f64 ┆ i64 ┆ i64 ┆ ┆ i64 ┆ i64 ┆ i64 ┆ i64 │\n", "╞════════╪══════════════╪══════════╪═════════╪═══╪═══════════════╪═══════════╪════════════╪════════╡\n", "│ male ┆ 4.0 ┆ 228 ┆ 0 ┆ … ┆ 4 ┆ 14 ┆ 5 ┆ 5 │\n", "│ male ┆ 7.0 ┆ 228 ┆ 0 ┆ … ┆ 4 ┆ 14 ┆ 5 ┆ 5 │\n", "│ male ┆ 15.0 ┆ 228 ┆ 0 ┆ … ┆ 4 ┆ 14 ┆ 5 ┆ 5 │\n", "│ female ┆ 4.0 ┆ 228 ┆ 0 ┆ … ┆ 4 ┆ 14 ┆ 5 ┆ 5 │\n", "│ female ┆ 7.0 ┆ 228 ┆ 0 ┆ … ┆ 4 ┆ 14 ┆ 5 ┆ 5 │\n", "│ female ┆ 15.0 ┆ 228 ┆ 0 ┆ … ┆ 4 ┆ 14 ┆ 5 ┆ 5 │\n", "└────────┴──────────────┴──────────┴─────────┴───┴───────────────┴───────────┴────────────┴────────┘" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "# New datagrid\n", "datagrid4 = me.datagrid(model4, \n", " gender=['male', 'female'],\n", " yearsmarried=quartiles) # can pass right in\n", "\n", "display(datagrid4)" ] }, { "cell_type": "markdown", "id": "8f88e6d8-7ef5-4ba4-9a8a-5cd78af0553d", "metadata": {}, "source": [ "And now we predict:" ] }, { "cell_type": "code", "execution_count": 21, "id": "659f0aa1-52af-4c23-8593-9a73dfe7aeee", "metadata": {}, "outputs": [ { "data": { "text/html": [ "
\n", "shape: (6, 18)
genderyearsmarriedrowidestimatestd_errorstatisticp_values_valueconf_lowconf_highrownamesaffairsagechildrenreligiousnesseducationoccupationrating
strf64i32f64f64f64f64f64f64f64i64i64f64stri64i64i64i64
"male"4.004.0570670.08059950.3364190.0inf3.8990964.215038228032.487521"yes"41455
"male"7.013.9647580.06509460.908150.0inf3.8371764.09234228032.487521"yes"41455
"male"15.023.71860.09909437.5259230.0inf3.5243793.912821228032.487521"yes"41455
"female"4.034.1938570.07403956.6436730.0inf4.0487424.338971228032.487521"yes"41455
"female"7.044.0040360.06120765.4180890.0inf3.8840734.123999228032.487521"yes"41455
"female"15.053.4978480.09607836.4064170.0inf3.3095393.686157228032.487521"yes"41455
" ], "text/plain": [ "shape: (6, 9)\n", "┌────────┬──────────────┬──────────┬───────────┬───┬─────────┬─────┬──────┬───────┐\n", "│ gender ┆ yearsmarried ┆ Estimate ┆ Std.Error ┆ … ┆ P(>|z|) ┆ S ┆ 2.5% ┆ 97.5% │\n", "│ --- ┆ --- ┆ --- ┆ --- ┆ ┆ --- ┆ --- ┆ --- ┆ --- │\n", "│ str ┆ str ┆ str ┆ str ┆ ┆ str ┆ str ┆ str ┆ str │\n", "╞════════╪══════════════╪══════════╪═══════════╪═══╪═════════╪═════╪══════╪═══════╡\n", "│ male ┆ 4 ┆ 4.06 ┆ 0.0806 ┆ … ┆ 0 ┆ inf ┆ 3.9 ┆ 4.22 │\n", "│ male ┆ 7 ┆ 3.96 ┆ 0.0651 ┆ … ┆ 0 ┆ inf ┆ 3.84 ┆ 4.09 │\n", "│ male ┆ 15 ┆ 3.72 ┆ 0.0991 ┆ … ┆ 0 ┆ inf ┆ 3.52 ┆ 3.91 │\n", "│ female ┆ 4 ┆ 4.19 ┆ 0.074 ┆ … ┆ 0 ┆ inf ┆ 4.05 ┆ 4.34 │\n", "│ female ┆ 7 ┆ 4 ┆ 0.0612 ┆ … ┆ 0 ┆ inf ┆ 3.88 ┆ 4.12 │\n", "│ female ┆ 15 ┆ 3.5 ┆ 0.0961 ┆ … ┆ 0 ┆ inf ┆ 3.31 ┆ 3.69 │\n", "└────────┴──────────────┴──────────┴───────────┴───┴─────────┴─────┴──────┴───────┘\n", "\n", "Columns: gender, yearsmarried, rowid, estimate, std_error, statistic, p_value, s_value, conf_low, conf_high, rownames, affairs, age, children, religiousness, education, occupation, rating" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "# Predict\n", "predictions4 = me.predictions(model4, newdata=datagrid4)\n", "display(predictions4)" ] }, { "cell_type": "markdown", "id": "fa80e00d-78f2-4127-b45a-2296d81d4355", "metadata": {}, "source": [ "We may choose to plot this for interpretability:" ] }, { "cell_type": "code", "execution_count": 22, "id": "aa08553e-9a35-4a4c-90b5-3dbc9247613b", "metadata": {}, "outputs": [ { "data": { "text/plain": [ "" ] }, "execution_count": 22, "metadata": {}, "output_type": "execute_result" }, { "data": { "image/png": "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", "text/plain": [ "
" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "# Plot\n", "sns.lineplot(data=predictions4,\n", " x='yearsmarried', y='estimate', \n", " hue='gender')" ] }, { "cell_type": "markdown", "id": "9a4718fb-4849-4b58-8d28-e83ac1ad812e", "metadata": {}, "source": [ "Interpreting the comparisons now proceeds as before. Depending on our questions, we might want to know the differences between men and women at each time point. We do this simply by specifying our hypotheses. Once again here are the predictions:" ] }, { "cell_type": "code", "execution_count": 23, "id": "15b6e4b7-a75f-40a9-b786-66a37631e0dc", "metadata": {}, "outputs": [ { "data": { "text/html": [ "
\n", "shape: (6, 18)
genderyearsmarriedrowidestimatestd_errorstatisticp_values_valueconf_lowconf_highrownamesaffairsagechildrenreligiousnesseducationoccupationrating
strf64i32f64f64f64f64f64f64f64i64i64f64stri64i64i64i64
"male"4.004.0570670.08059950.3364190.0inf3.8990964.215038228032.487521"yes"41455
"male"7.013.9647580.06509460.908150.0inf3.8371764.09234228032.487521"yes"41455
"male"15.023.71860.09909437.5259230.0inf3.5243793.912821228032.487521"yes"41455
"female"4.034.1938570.07403956.6436730.0inf4.0487424.338971228032.487521"yes"41455
"female"7.044.0040360.06120765.4180890.0inf3.8840734.123999228032.487521"yes"41455
"female"15.053.4978480.09607836.4064170.0inf3.3095393.686157228032.487521"yes"41455
" ], "text/plain": [ "shape: (6, 9)\n", "┌────────┬──────────────┬──────────┬───────────┬───┬─────────┬─────┬──────┬───────┐\n", "│ gender ┆ yearsmarried ┆ Estimate ┆ Std.Error ┆ … ┆ P(>|z|) ┆ S ┆ 2.5% ┆ 97.5% │\n", "│ --- ┆ --- ┆ --- ┆ --- ┆ ┆ --- ┆ --- ┆ --- ┆ --- │\n", "│ str ┆ str ┆ str ┆ str ┆ ┆ str ┆ str ┆ str ┆ str │\n", "╞════════╪══════════════╪══════════╪═══════════╪═══╪═════════╪═════╪══════╪═══════╡\n", "│ male ┆ 4 ┆ 4.06 ┆ 0.0806 ┆ … ┆ 0 ┆ inf ┆ 3.9 ┆ 4.22 │\n", "│ male ┆ 7 ┆ 3.96 ┆ 0.0651 ┆ … ┆ 0 ┆ inf ┆ 3.84 ┆ 4.09 │\n", "│ male ┆ 15 ┆ 3.72 ┆ 0.0991 ┆ … ┆ 0 ┆ inf ┆ 3.52 ┆ 3.91 │\n", "│ female ┆ 4 ┆ 4.19 ┆ 0.074 ┆ … ┆ 0 ┆ inf ┆ 4.05 ┆ 4.34 │\n", "│ female ┆ 7 ┆ 4 ┆ 0.0612 ┆ … ┆ 0 ┆ inf ┆ 3.88 ┆ 4.12 │\n", "│ female ┆ 15 ┆ 3.5 ┆ 0.0961 ┆ … ┆ 0 ┆ inf ┆ 3.31 ┆ 3.69 │\n", "└────────┴──────────────┴──────────┴───────────┴───┴─────────┴─────┴──────┴───────┘\n", "\n", "Columns: gender, yearsmarried, rowid, estimate, std_error, statistic, p_value, s_value, conf_low, conf_high, rownames, affairs, age, children, religiousness, education, occupation, rating" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "# Predictions\n", "display(predictions4)" ] }, { "cell_type": "markdown", "id": "bf5bb614-59dd-4d99-ad1b-d2a665d37ab5", "metadata": {}, "source": [ "Getting all the comparisons done in one go is a little clunky. We can do it one at a time, e.g:\n", "\n", "`me.predictions(model4, newdata=datagrid4, hypothesis='b1=b4')`\n", "\n", "or we could recall which rows we wanted to find the differences of (e.g. row 1 - row 4 compares men and women at 4 years of marriage, and ask for all the pairwise comparisons:\n" ] }, { "cell_type": "code", "execution_count": 24, "id": "ea78dbc6-35d0-4550-9b58-94b8a4011389", "metadata": {}, "outputs": [ { "data": { "text/html": [ "
\n", "\n", "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
termestimatestd_errorstatisticp_values_valueconf_lowconf_high
0Row 1 - Row 20.0923090.0344532.6792727.378243e-037.0825070.0247820.159836
1Row 1 - Row 30.3384670.1263282.6792727.378245e-037.0825070.0908690.586065
2Row 1 - Row 4-0.1367900.109444-1.2498612.113505e-012.242291-0.3512960.077717
3Row 1 - Row 50.0530310.1012050.5239926.002839e-010.736283-0.1453280.251389
4Row 1 - Row 60.5592190.1254084.4592018.226583e-0616.8912750.3134240.805014
5Row 2 - Row 30.2461580.0918752.6792727.378245e-037.0825070.0660860.426229
6Row 2 - Row 4-0.2290990.098585-2.3238672.013263e-025.634321-0.422323-0.035875
7Row 2 - Row 5-0.0392780.089351-0.4395986.602280e-010.598964-0.2144020.135845
8Row 2 - Row 60.4669100.1160524.0232675.739648e-0514.0886780.2394510.694369
9Row 3 - Row 4-0.4752570.123699-3.8420371.220176e-0413.000623-0.717702-0.232811
10Row 3 - Row 5-0.2854360.116473-2.4506641.425932e-026.131952-0.513719-0.057153
11Row 3 - Row 60.2207520.1380241.5993791.097365e-013.187885-0.0497690.491274
12Row 4 - Row 50.1898210.0321605.9024033.582456e-0928.0564040.1267880.252853
13Row 4 - Row 60.6960090.1179205.9024033.582456e-0928.0564040.4648910.927127
14Row 5 - Row 60.5061880.0857605.9024033.582456e-0928.0564040.3381020.674274
\n", "
" ], "text/plain": [ " term estimate std_error statistic p_value s_value \\\n", "0 Row 1 - Row 2 0.092309 0.034453 2.679272 7.378243e-03 7.082507 \n", "1 Row 1 - Row 3 0.338467 0.126328 2.679272 7.378245e-03 7.082507 \n", "2 Row 1 - Row 4 -0.136790 0.109444 -1.249861 2.113505e-01 2.242291 \n", "3 Row 1 - Row 5 0.053031 0.101205 0.523992 6.002839e-01 0.736283 \n", "4 Row 1 - Row 6 0.559219 0.125408 4.459201 8.226583e-06 16.891275 \n", "5 Row 2 - Row 3 0.246158 0.091875 2.679272 7.378245e-03 7.082507 \n", "6 Row 2 - Row 4 -0.229099 0.098585 -2.323867 2.013263e-02 5.634321 \n", "7 Row 2 - Row 5 -0.039278 0.089351 -0.439598 6.602280e-01 0.598964 \n", "8 Row 2 - Row 6 0.466910 0.116052 4.023267 5.739648e-05 14.088678 \n", "9 Row 3 - Row 4 -0.475257 0.123699 -3.842037 1.220176e-04 13.000623 \n", "10 Row 3 - Row 5 -0.285436 0.116473 -2.450664 1.425932e-02 6.131952 \n", "11 Row 3 - Row 6 0.220752 0.138024 1.599379 1.097365e-01 3.187885 \n", "12 Row 4 - Row 5 0.189821 0.032160 5.902403 3.582456e-09 28.056404 \n", "13 Row 4 - Row 6 0.696009 0.117920 5.902403 3.582456e-09 28.056404 \n", "14 Row 5 - Row 6 0.506188 0.085760 5.902403 3.582456e-09 28.056404 \n", "\n", " conf_low conf_high \n", "0 0.024782 0.159836 \n", "1 0.090869 0.586065 \n", "2 -0.351296 0.077717 \n", "3 -0.145328 0.251389 \n", "4 0.313424 0.805014 \n", "5 0.066086 0.426229 \n", "6 -0.422323 -0.035875 \n", "7 -0.214402 0.135845 \n", "8 0.239451 0.694369 \n", "9 -0.717702 -0.232811 \n", "10 -0.513719 -0.057153 \n", "11 -0.049769 0.491274 \n", "12 0.126788 0.252853 \n", "13 0.464891 0.927127 \n", "14 0.338102 0.674274 " ] }, "execution_count": 24, "metadata": {}, "output_type": "execute_result" } ], "source": [ "# Clunky comparisons\n", "me.predictions(model4, \n", " newdata=datagrid4,\n", " hypothesis='pairwise').to_pandas()" ] }, { "cell_type": "markdown", "id": "1af41d99-98df-4c7d-8d25-d1b7d79147ca", "metadata": {}, "source": [ "This sort of approach is sometimes useful, but a cleaner technique is not to dwell on the individual predictions, but rather to look at the simple slopes!" ] }, { "cell_type": "markdown", "id": "99988025-6d26-4627-886d-9f5ebd34c657", "metadata": {}, "source": [ "#### Exploring more interactions - categorical and continuous - **slopes** \n", "When dealing with situations like this, its almost always better to focus instead on the simple slopes. These represent, quite literally, the slope the continuous variable has at *the level of another variable*, which might be continuous or categorical itself. Its akin to asking this question:\n", "\n", "- How much does Y change with a 1-unit increase in X, when Z is fixed to a specific category?\n", "Or, here:\n", "- How much does marital satisfaction change when years married increases by 1 year, when gender is fixed to male or female?\n", "\n", "This is a much cleaner approach as it removes the need to make lots of comparisons and instead directly checks the associations of a predictor and the dependent variable, fixing one variable at specific value. Lets see how this works, using the `me.slopes` function. First, examining the slopes visually makes it clear something is going on:" ] }, { "cell_type": "code", "execution_count": 25, "id": "65201499-f7f6-4238-b93a-ec46226b92fb", "metadata": {}, "outputs": [ { "data": { "text/plain": [ "" ] }, "execution_count": 25, "metadata": {}, "output_type": "execute_result" }, { "data": { "image/png": "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", "text/plain": [ "
" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "# Plot\n", "sns.lineplot(data=predictions4,\n", " x='yearsmarried', y='estimate', \n", " hue='gender')" ] }, { "cell_type": "markdown", "id": "18ae853f-8e43-4eec-b71a-f76fb9aa7542", "metadata": {}, "source": [ "Lets use `me.slopes` to work those out. Notice here we don't actually specify a new data grid, but simply say which variable to figure out the slopes for (`yearsmarried` in this case), and which variable to do this over (`gender`)." ] }, { "cell_type": "code", "execution_count": 26, "id": "5134e532-8950-4d41-a795-13af353f43ba", "metadata": {}, "outputs": [ { "data": { "text/html": [ "
\n", "shape: (2, 10)
gendertermcontrastestimatestd_errorstatisticp_values_valueconf_lowconf_high
strstrstrf64f64f64f64f64f64f64
"female""yearsmarried""mean(dY/dX)"-0.0632740.010686-5.9211643.1967e-928.220764-0.084218-0.042329
"male""yearsmarried""mean(dY/dX)"-0.030770.011479-2.6805260.0073517.087913-0.053268-0.008271
" ], "text/plain": [ "shape: (2, 10)\n", "┌────────┬──────────────┬─────────────┬──────────┬───┬─────────┬──────┬─────────┬──────────┐\n", "│ gender ┆ Term ┆ Contrast ┆ Estimate ┆ … ┆ P(>|z|) ┆ S ┆ 2.5% ┆ 97.5% │\n", "│ --- ┆ --- ┆ --- ┆ --- ┆ ┆ --- ┆ --- ┆ --- ┆ --- │\n", "│ str ┆ str ┆ str ┆ str ┆ ┆ str ┆ str ┆ str ┆ str │\n", "╞════════╪══════════════╪═════════════╪══════════╪═══╪═════════╪══════╪═════════╪══════════╡\n", "│ female ┆ yearsmarried ┆ mean(dY/dX) ┆ -0.0633 ┆ … ┆ 3.2e-09 ┆ 28.2 ┆ -0.0842 ┆ -0.0423 │\n", "│ male ┆ yearsmarried ┆ mean(dY/dX) ┆ -0.0308 ┆ … ┆ 0.00735 ┆ 7.09 ┆ -0.0533 ┆ -0.00827 │\n", "└────────┴──────────────┴─────────────┴──────────┴───┴─────────┴──────┴─────────┴──────────┘\n", "\n", "Columns: gender, term, contrast, estimate, std_error, statistic, p_value, s_value, conf_low, conf_high" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "# slopes\n", "slopes = me.slopes(model4, \n", " variables='yearsmarried', \n", " by='gender')\n", "display(slopes)" ] }, { "cell_type": "markdown", "id": "7650fb88-a29b-4a13-a7d4-df1218facf5d", "metadata": {}, "source": [ "Aligned with the figure, females show a steeper decline than males (a more negative slope). We can ask whether there is a difference between those slopes in the usual way, e.g." ] }, { "cell_type": "code", "execution_count": 27, "id": "baefeb65-0fe5-4f1d-9095-b72d05db7609", "metadata": {}, "outputs": [ { "data": { "text/html": [ "
\n", "shape: (1, 8)
termestimatestd_errorstatisticp_values_valueconf_lowconf_high
strf64f64f64f64f64f64f64
"Row 1 - Row 2"-0.0325040.015699-2.0704940.0384064.702519-0.063273-0.001735
" ], "text/plain": [ "shape: (1, 8)\n", "┌───────────────┬──────────┬───────────┬───────┬─────────┬─────┬─────────┬──────────┐\n", "│ Term ┆ Estimate ┆ Std.Error ┆ z ┆ P(>|z|) ┆ S ┆ 2.5% ┆ 97.5% │\n", "│ --- ┆ --- ┆ --- ┆ --- ┆ --- ┆ --- ┆ --- ┆ --- │\n", "│ str ┆ str ┆ str ┆ str ┆ str ┆ str ┆ str ┆ str │\n", "╞═══════════════╪══════════╪═══════════╪═══════╪═════════╪═════╪═════════╪══════════╡\n", "│ Row 1 - Row 2 ┆ -0.0325 ┆ 0.0157 ┆ -2.07 ┆ 0.0384 ┆ 4.7 ┆ -0.0633 ┆ -0.00174 │\n", "└───────────────┴──────────┴───────────┴───────┴─────────┴─────┴─────────┴──────────┘\n", "\n", "Columns: term, estimate, std_error, statistic, p_value, s_value, conf_low, conf_high" ] }, "execution_count": 27, "metadata": {}, "output_type": "execute_result" } ], "source": [ "# Test for differences therein\n", "me.slopes(model4, \n", " variables='yearsmarried', \n", " by='gender',\n", " hypothesis='pairwise')" ] }, { "cell_type": "markdown", "id": "150bb983-8a94-4f31-80b6-22f2a4802b38", "metadata": {}, "source": [ "### Exploring more interactions - continuous and continuous!\n" ] }, { "cell_type": "markdown", "id": "b161f730-3ba3-4b6b-a06a-cd361c583968", "metadata": {}, "source": [ "Let us now consider a model that has an interaction between two continuous predictors, and how marginal predictions can help us interpret them. Here, we build a model that predicts marital satisfaction from number of years married and education. We are interested in whether those with more education and longer marriages are happier than those with less education.\n", "\n", "First we fit the model:" ] }, { "cell_type": "code", "execution_count": 28, "id": "bcaa88fc-4c46-44e1-bae1-ab3308f1cd7a", "metadata": {}, "outputs": [ { "data": { "text/html": [ "\n", "\n", "\n", " \n", "\n", "\n", " \n", "\n", "\n", " \n", "\n", "\n", " \n", "\n", "
OLS Regression Results
Dep. Variable: rating R-squared: 0.097
Model: OLS Adj. R-squared: 0.092
No. Observations: 601 F-statistic: 21.28
Covariance Type: nonrobust Prob (F-statistic): 4.17e-13
\n", "\n", "\n", " \n", "\n", "\n", " \n", "\n", "\n", " \n", "\n", "\n", " \n", "\n", "\n", " \n", "\n", "
coef std err t P>|t| [0.025 0.975]
Intercept 5.4446 0.589 9.246 0.000 4.288 6.601
yearsmarried -0.2574 0.054 -4.798 0.000 -0.363 -0.152
education -0.0700 0.036 -1.919 0.056 -0.142 0.002
yearsmarried:education 0.0130 0.003 3.924 0.000 0.006 0.019


Notes:
[1] Standard Errors assume that the covariance matrix of the errors is correctly specified.
[2] The condition number is large, 2.26e+03. This might indicate that there are
strong multicollinearity or other numerical problems." ], "text/latex": [ "\\begin{center}\n", "\\begin{tabular}{lclc}\n", "\\toprule\n", "\\textbf{Dep. Variable:} & rating & \\textbf{ R-squared: } & 0.097 \\\\\n", "\\textbf{Model:} & OLS & \\textbf{ Adj. R-squared: } & 0.092 \\\\\n", "\\textbf{No. Observations:} & 601 & \\textbf{ F-statistic: } & 21.28 \\\\\n", "\\textbf{Covariance Type:} & nonrobust & \\textbf{ Prob (F-statistic):} & 4.17e-13 \\\\\n", "\\bottomrule\n", "\\end{tabular}\n", "\\begin{tabular}{lcccccc}\n", " & \\textbf{coef} & \\textbf{std err} & \\textbf{t} & \\textbf{P$> |$t$|$} & \\textbf{[0.025} & \\textbf{0.975]} \\\\\n", "\\midrule\n", "\\textbf{Intercept} & 5.4446 & 0.589 & 9.246 & 0.000 & 4.288 & 6.601 \\\\\n", "\\textbf{yearsmarried} & -0.2574 & 0.054 & -4.798 & 0.000 & -0.363 & -0.152 \\\\\n", "\\textbf{education} & -0.0700 & 0.036 & -1.919 & 0.056 & -0.142 & 0.002 \\\\\n", "\\textbf{yearsmarried:education} & 0.0130 & 0.003 & 3.924 & 0.000 & 0.006 & 0.019 \\\\\n", "\\bottomrule\n", "\\end{tabular}\n", "%\\caption{OLS Regression Results}\n", "\\end{center}\n", "\n", "Notes: \\newline\n", " [1] Standard Errors assume that the covariance matrix of the errors is correctly specified. \\newline\n", " [2] The condition number is large, 2.26e+03. This might indicate that there are \\newline\n", " strong multicollinearity or other numerical problems." ], "text/plain": [ "\n", "\"\"\"\n", " OLS Regression Results \n", "==============================================================================\n", "Dep. Variable: rating R-squared: 0.097\n", "Model: OLS Adj. R-squared: 0.092\n", "No. Observations: 601 F-statistic: 21.28\n", "Covariance Type: nonrobust Prob (F-statistic): 4.17e-13\n", "==========================================================================================\n", " coef std err t P>|t| [0.025 0.975]\n", "------------------------------------------------------------------------------------------\n", "Intercept 5.4446 0.589 9.246 0.000 4.288 6.601\n", "yearsmarried -0.2574 0.054 -4.798 0.000 -0.363 -0.152\n", "education -0.0700 0.036 -1.919 0.056 -0.142 0.002\n", "yearsmarried:education 0.0130 0.003 3.924 0.000 0.006 0.019\n", "==========================================================================================\n", "\n", "Notes:\n", "[1] Standard Errors assume that the covariance matrix of the errors is correctly specified.\n", "[2] The condition number is large, 2.26e+03. This might indicate that there are\n", "strong multicollinearity or other numerical problems.\n", "\"\"\"" ] }, "execution_count": 28, "metadata": {}, "output_type": "execute_result" } ], "source": [ "# Model 5\n", "model5 = smf.ols('rating ~ yearsmarried * education', data=affairs).fit()\n", "model5.summary(slim=True)" ] }, { "cell_type": "markdown", "id": "cfd7578a-0550-4af8-8780-0045958814d9", "metadata": {}, "source": [ "How should we proceed in terms of inference, given we can see the interaction effect is significant?\n", "\n", "Much like before, we can select some points on which to compare our variables, and try to interpret the differences between them.\n", "\n", "Here, I've chosen to predict for marriages of lengths 1, 5, 10, and 20 years, and educational attainment of 10, 12, 18, and 20. " ] }, { "cell_type": "code", "execution_count": 29, "id": "04d8009c-5015-4d49-9e49-b938d56f97c9", "metadata": {}, "outputs": [], "source": [ "# Create a datagrid\n", "datagrid5 = me.datagrid(model5,\n", " yearsmarried=[1, 5, 10, 20],\n", " education=[10, 12, 18, 20])\n", "\n", "# Generate predictions\n", "predictions5 = me.predictions(model5, newdata=datagrid5)" ] }, { "cell_type": "markdown", "id": "34155e1e-5df1-4709-a188-42a8c4d55455", "metadata": {}, "source": [ "Plotting these predictions will help visualise what's going on." ] }, { "cell_type": "code", "execution_count": 30, "id": "5f14d43f-5a11-4f68-a2b9-3fd796bcdc9c", "metadata": {}, "outputs": [ { "data": { "text/plain": [ "" ] }, "execution_count": 30, "metadata": {}, "output_type": "execute_result" }, { "data": { "image/png": "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", "text/plain": [ "
" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "# Plot\n", "sns.lineplot(data=predictions5, \n", " x='yearsmarried', y='estimate',\n", " hue='education')" ] }, { "cell_type": "markdown", "id": "d484fc75-333d-4346-813a-e4b0baac2dbc", "metadata": {}, "source": [ "It is difficult to know where to begin! It is often simpler in these instances to pick one or two points and test differences. \n", "\n", "Lets say that we want to know whether the difference between educational attainment values 12 and 18 are significantly different at 10 and 20 years of marriage. That is achievable:" ] }, { "cell_type": "code", "execution_count": 31, "id": "af1d3b3b-a5bd-4350-b832-1fc00dc8aa1f", "metadata": {}, "outputs": [ { "data": { "text/html": [ "
\n", "shape: (4, 18)
yearsmarriededucationrowidestimatestd_errorstatisticp_values_valueconf_lowconf_highrownamesaffairsgenderagechildrenreligiousnessoccupationrating
i64i64i32f64f64f64f64f64f64f64i64i64strf64stri64i64i64
101203.5887390.08780140.8734590.0inf3.4166523.76082714180"female"32.487521"yes"455
101813.9479710.05545571.1923190.0inf3.8392814.05666114180"female"32.487521"yes"455
201222.5723240.18963413.5646470.0inf2.2006482.94400114180"female"32.487521"yes"455
201833.710520.12379629.9727650.0inf3.4678843.95315614180"female"32.487521"yes"455
" ], "text/plain": [ "shape: (4, 9)\n", "┌──────────────┬───────────┬──────────┬───────────┬───┬─────────┬─────┬──────┬───────┐\n", "│ yearsmarried ┆ education ┆ Estimate ┆ Std.Error ┆ … ┆ P(>|z|) ┆ S ┆ 2.5% ┆ 97.5% │\n", "│ --- ┆ --- ┆ --- ┆ --- ┆ ┆ --- ┆ --- ┆ --- ┆ --- │\n", "│ str ┆ str ┆ str ┆ str ┆ ┆ str ┆ str ┆ str ┆ str │\n", "╞══════════════╪═══════════╪══════════╪═══════════╪═══╪═════════╪═════╪══════╪═══════╡\n", "│ 10 ┆ 12 ┆ 3.59 ┆ 0.0878 ┆ … ┆ 0 ┆ inf ┆ 3.42 ┆ 3.76 │\n", "│ 10 ┆ 18 ┆ 3.95 ┆ 0.0555 ┆ … ┆ 0 ┆ inf ┆ 3.84 ┆ 4.06 │\n", "│ 20 ┆ 12 ┆ 2.57 ┆ 0.19 ┆ … ┆ 0 ┆ inf ┆ 2.2 ┆ 2.94 │\n", "│ 20 ┆ 18 ┆ 3.71 ┆ 0.124 ┆ … ┆ 0 ┆ inf ┆ 3.47 ┆ 3.95 │\n", "└──────────────┴───────────┴──────────┴───────────┴───┴─────────┴─────┴──────┴───────┘\n", "\n", "Columns: yearsmarried, education, rowid, estimate, std_error, statistic, p_value, s_value, conf_low, conf_high, rownames, affairs, gender, age, children, religiousness, occupation, rating" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "# Sensible predictions\n", "display(me.predictions(model5, \n", " newdata=me.datagrid(model5, yearsmarried=[10, 20], education=[12, 18])\n", " )\n", " )" ] }, { "cell_type": "markdown", "id": "36151c6a-6f55-4aa1-b790-b87de264b292", "metadata": {}, "source": [ "Repeat and actually test difference between rows 1 and 2 (married = 10, education varies):" ] }, { "cell_type": "code", "execution_count": 32, "id": "22085f91-e48b-48a0-a265-13b5b2d7f371", "metadata": {}, "outputs": [ { "data": { "text/html": [ "
\n", "shape: (1, 8)
termestimatestd_errorstatisticp_values_valueconf_lowconf_high
strf64f64f64f64f64f64f64
"b2=b1"0.3592310.1075443.3403320.00083710.2228620.148450.570013
" ], "text/plain": [ "shape: (1, 8)\n", "┌───────┬──────────┬───────────┬──────┬──────────┬──────┬───────┬───────┐\n", "│ Term ┆ Estimate ┆ Std.Error ┆ z ┆ P(>|z|) ┆ S ┆ 2.5% ┆ 97.5% │\n", "│ --- ┆ --- ┆ --- ┆ --- ┆ --- ┆ --- ┆ --- ┆ --- │\n", "│ str ┆ str ┆ str ┆ str ┆ str ┆ str ┆ str ┆ str │\n", "╞═══════╪══════════╪═══════════╪══════╪══════════╪══════╪═══════╪═══════╡\n", "│ b2=b1 ┆ 0.359 ┆ 0.108 ┆ 3.34 ┆ 0.000837 ┆ 10.2 ┆ 0.148 ┆ 0.57 │\n", "└───────┴──────────┴───────────┴──────┴──────────┴──────┴───────┴───────┘\n", "\n", "Columns: term, estimate, std_error, statistic, p_value, s_value, conf_low, conf_high" ] }, "execution_count": 32, "metadata": {}, "output_type": "execute_result" } ], "source": [ "# Testing one specific hypothesis\n", "me.predictions(model5, \n", " newdata=me.datagrid(model5, yearsmarried=[10], education=[12, 18]),\n", " hypothesis='b2=b1')" ] }, { "cell_type": "code", "execution_count": 33, "id": "85afae59-d350-4d64-a911-736d82e70679", "metadata": {}, "outputs": [ { "data": { "text/html": [ "
\n", "shape: (1, 8)
termestimatestd_errorstatisticp_values_valueconf_lowconf_high
strf64f64f64f64f64f64f64
"b2=b1"1.1381960.2325814.8937619.8927e-719.9471320.6823451.594046
" ], "text/plain": [ "shape: (1, 8)\n", "┌───────┬──────────┬───────────┬──────┬──────────┬──────┬───────┬───────┐\n", "│ Term ┆ Estimate ┆ Std.Error ┆ z ┆ P(>|z|) ┆ S ┆ 2.5% ┆ 97.5% │\n", "│ --- ┆ --- ┆ --- ┆ --- ┆ --- ┆ --- ┆ --- ┆ --- │\n", "│ str ┆ str ┆ str ┆ str ┆ str ┆ str ┆ str ┆ str │\n", "╞═══════╪══════════╪═══════════╪══════╪══════════╪══════╪═══════╪═══════╡\n", "│ b2=b1 ┆ 1.14 ┆ 0.233 ┆ 4.89 ┆ 9.89e-07 ┆ 19.9 ┆ 0.682 ┆ 1.59 │\n", "└───────┴──────────┴───────────┴──────┴──────────┴──────┴───────┴───────┘\n", "\n", "Columns: term, estimate, std_error, statistic, p_value, s_value, conf_low, conf_high" ] }, "execution_count": 33, "metadata": {}, "output_type": "execute_result" } ], "source": [ "# And repeats for another\n", "me.predictions(model5, \n", " newdata=me.datagrid(model5, yearsmarried=[20], education=[12, 18]),\n", " hypothesis='b2=b1')" ] }, { "cell_type": "markdown", "id": "c668ec49-6477-44bf-8d14-443f09e42aef", "metadata": {}, "source": [ "Clearly a bigger difference at 20 years between the low and high education groups...\n", "\n", "But what if you want to compare whether the *difference at 20 years between the education values* is *significantly bigger* than the difference *at 10 years?* Thats a neccessary and complex question, but you can ask it by actually performing the calculation with the `hypothesis` argument on the full set of comparisons. Lets see all the necessary predictions again first:" ] }, { "cell_type": "code", "execution_count": 34, "id": "261b09b5-a274-4932-b790-90eb9ba483ce", "metadata": {}, "outputs": [ { "data": { "text/html": [ "
\n", "shape: (4, 18)
yearsmarriededucationrowidestimatestd_errorstatisticp_values_valueconf_lowconf_highrownamesaffairsgenderagechildrenreligiousnessoccupationrating
i64i64i32f64f64f64f64f64f64f64i64i64strf64stri64i64i64
101203.5887390.08780140.8734590.0inf3.4166523.76082719300"female"32.487521"yes"455
101813.9479710.05545571.1923190.0inf3.8392814.05666119300"female"32.487521"yes"455
201222.5723240.18963413.5646470.0inf2.2006482.94400119300"female"32.487521"yes"455
201833.710520.12379629.9727650.0inf3.4678843.95315619300"female"32.487521"yes"455
" ], "text/plain": [ "shape: (4, 9)\n", "┌──────────────┬───────────┬──────────┬───────────┬───┬─────────┬─────┬──────┬───────┐\n", "│ yearsmarried ┆ education ┆ Estimate ┆ Std.Error ┆ … ┆ P(>|z|) ┆ S ┆ 2.5% ┆ 97.5% │\n", "│ --- ┆ --- ┆ --- ┆ --- ┆ ┆ --- ┆ --- ┆ --- ┆ --- │\n", "│ str ┆ str ┆ str ┆ str ┆ ┆ str ┆ str ┆ str ┆ str │\n", "╞══════════════╪═══════════╪══════════╪═══════════╪═══╪═════════╪═════╪══════╪═══════╡\n", "│ 10 ┆ 12 ┆ 3.59 ┆ 0.0878 ┆ … ┆ 0 ┆ inf ┆ 3.42 ┆ 3.76 │\n", "│ 10 ┆ 18 ┆ 3.95 ┆ 0.0555 ┆ … ┆ 0 ┆ inf ┆ 3.84 ┆ 4.06 │\n", "│ 20 ┆ 12 ┆ 2.57 ┆ 0.19 ┆ … ┆ 0 ┆ inf ┆ 2.2 ┆ 2.94 │\n", "│ 20 ┆ 18 ┆ 3.71 ┆ 0.124 ┆ … ┆ 0 ┆ inf ┆ 3.47 ┆ 3.95 │\n", "└──────────────┴───────────┴──────────┴───────────┴───┴─────────┴─────┴──────┴───────┘\n", "\n", "Columns: yearsmarried, education, rowid, estimate, std_error, statistic, p_value, s_value, conf_low, conf_high, rownames, affairs, gender, age, children, religiousness, occupation, rating" ] }, "execution_count": 34, "metadata": {}, "output_type": "execute_result" } ], "source": [ "# These are the predictions\n", "me.predictions(model5, \n", " newdata=me.datagrid(model5,\n", " yearsmarried=[10, 20],\n", " education=[12, 18])\n", " )" ] }, { "cell_type": "markdown", "id": "ce6fa121-791a-4df7-b413-930c5b7ade73", "metadata": {}, "source": [ "We specifically want to see whether the difference between rows 3 and 4 (years married = 20, education 12 and 18) is bigger than rows 1 and 2 (years married = 10, education 12 and 18). We can do that literally translating that query into a hypothesis." ] }, { "cell_type": "code", "execution_count": 35, "id": "f77f1b30-b47d-4bfc-b087-58feca38a521", "metadata": {}, "outputs": [ { "data": { "text/html": [ "
\n", "shape: (1, 8)
termestimatestd_errorstatisticp_values_valueconf_lowconf_high
strf64f64f64f64f64f64f64
"b4-b3=b2-b1"0.7789640.1985293.9236890.00008713.4852590.3898551.168073
" ], "text/plain": [ "shape: (1, 8)\n", "┌─────────────┬──────────┬───────────┬──────┬──────────┬──────┬──────┬───────┐\n", "│ Term ┆ Estimate ┆ Std.Error ┆ z ┆ P(>|z|) ┆ S ┆ 2.5% ┆ 97.5% │\n", "│ --- ┆ --- ┆ --- ┆ --- ┆ --- ┆ --- ┆ --- ┆ --- │\n", "│ str ┆ str ┆ str ┆ str ┆ str ┆ str ┆ str ┆ str │\n", "╞═════════════╪══════════╪═══════════╪══════╪══════════╪══════╪══════╪═══════╡\n", "│ b4-b3=b2-b1 ┆ 0.779 ┆ 0.199 ┆ 3.92 ┆ 8.72e-05 ┆ 13.5 ┆ 0.39 ┆ 1.17 │\n", "└─────────────┴──────────┴───────────┴──────┴──────────┴──────┴──────┴───────┘\n", "\n", "Columns: term, estimate, std_error, statistic, p_value, s_value, conf_low, conf_high" ] }, "execution_count": 35, "metadata": {}, "output_type": "execute_result" } ], "source": [ "# Answer the question!\n", "me.predictions(model5, \n", " newdata=me.datagrid(model5,\n", " yearsmarried=[10, 20],\n", " education=[12, 18]),\n", " hypothesis='b4-b3 = b2-b1'\n", " )" ] }, { "cell_type": "markdown", "id": "c27b27b4-22b5-495f-957f-059118f23cf0", "metadata": {}, "source": [ "This is indeed significant and reveals a closer insight into what the model understands. \n", "\n", "Of course, another approach is to compare slopes. We saw how to do this earlier in a continuous by categorical interaction situation, but here, everything is continuous.\n", "\n", "The solution is essentially a combination of using a datagrid and the `slopes` command. We pick some candidate values for our predictors, generate a datagrid, and then compute the slopes for one over the other variable. \n", "\n", "To be explicit, first we will plot the slopes. I want the slopes for three levels of education - 9, 14, and 20. Call these low, medium, and high. I want these slopes for as many years married as we care to consider as we will be averaging over it soon enough. This looks like the below:" ] }, { "cell_type": "code", "execution_count": 36, "id": "b19e37aa-1320-45f3-b3a8-8f4950f14128", "metadata": {}, "outputs": [ { "data": { "text/plain": [ "" ] }, "execution_count": 36, "metadata": {}, "output_type": "execute_result" }, { "data": { "image/png": "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", "text/plain": [ "
" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "# Get our datagrid\n", "slopes_datagrid = me.datagrid(model5, \n", " yearsmarried=[0.5, 1, 2, 3, 5, 10, 20],\n", " education=[9, 14, 20])\n", "\n", "# Predict them\n", "p = me.predictions(model5, newdata=slopes_datagrid)\n", "\n", "# Plot them\n", "sns.lineplot(data=p,\n", " x='yearsmarried', y='estimate',\n", " hue='education')" ] }, { "cell_type": "markdown", "id": "fd6b9a2d-8c7e-40b1-a59f-b8ecc6f2eb99", "metadata": {}, "source": [ "What do we expect? The slope for an educational attainment of 20 is basically zero - a flat line! For 14, it is negative but not too steep, but for 9, it is very steep and negative. So lets obtain those three values and see how they look, which we do with a combination of our existing data grid and the variables/by arguments you've already seen:" ] }, { "cell_type": "code", "execution_count": 37, "id": "6a14d692-53d0-44db-9c33-4faab39fa1fc", "metadata": {}, "outputs": [ { "data": { "text/html": [ "
\n", "shape: (3, 10)
educationtermcontrastestimatestd_errorstatisticp_values_valueconf_lowconf_high
i64strstrf64f64f64f64f64f64f64
9"yearsmarried""mean(dY/dX)"-0.140590.024533-5.7306991.0002e-826.57517-0.188673-0.092506
14"yearsmarried""mean(dY/dX)"-0.0756760.010247-7.3853231.5210e-1342.58004-0.095759-0.055593
20"yearsmarried""mean(dY/dX)"0.002220.0151430.1466240.8834290.178814-0.027460.031901
" ], "text/plain": [ "shape: (3, 10)\n", "┌───────────┬──────────────┬─────────────┬──────────┬───┬──────────┬───────┬─────────┬─────────┐\n", "│ education ┆ Term ┆ Contrast ┆ Estimate ┆ … ┆ P(>|z|) ┆ S ┆ 2.5% ┆ 97.5% │\n", "│ --- ┆ --- ┆ --- ┆ --- ┆ ┆ --- ┆ --- ┆ --- ┆ --- │\n", "│ str ┆ str ┆ str ┆ str ┆ ┆ str ┆ str ┆ str ┆ str │\n", "╞═══════════╪══════════════╪═════════════╪══════════╪═══╪══════════╪═══════╪═════════╪═════════╡\n", "│ 9 ┆ yearsmarried ┆ mean(dY/dX) ┆ -0.141 ┆ … ┆ 1e-08 ┆ 26.6 ┆ -0.189 ┆ -0.0925 │\n", "│ 14 ┆ yearsmarried ┆ mean(dY/dX) ┆ -0.0757 ┆ … ┆ 1.52e-13 ┆ 42.6 ┆ -0.0958 ┆ -0.0556 │\n", "│ 20 ┆ yearsmarried ┆ mean(dY/dX) ┆ 0.00222 ┆ … ┆ 0.883 ┆ 0.179 ┆ -0.0275 ┆ 0.0319 │\n", "└───────────┴──────────────┴─────────────┴──────────┴───┴──────────┴───────┴─────────┴─────────┘\n", "\n", "Columns: education, term, contrast, estimate, std_error, statistic, p_value, s_value, conf_low, conf_high" ] }, "execution_count": 37, "metadata": {}, "output_type": "execute_result" } ], "source": [ "# Slopes for ultimate insight\n", "final_slopes = me.slopes(model5,\n", " newdata=slopes_datagrid,\n", " variables='yearsmarried', # average over this\n", " by='education') # for each of these\n", "final_slopes" ] }, { "cell_type": "markdown", "id": "6f107c33-5016-44af-9dbe-8b97e7967941", "metadata": {}, "source": [ "The slope for an educational attainment of 20 is not significantly different from zero. As we predicted, for 14 its negative, and for 9, its much more steep. We can infact compare those if we want to using `hypothesis` - e.g - does the jump from 9 to 14 really slow your decline in marital satisfaction?" ] }, { "cell_type": "code", "execution_count": 38, "id": "a7bcb038-3484-4e27-8ad6-b90ce71e1b47", "metadata": {}, "outputs": [ { "data": { "text/html": [ "
\n", "shape: (1, 8)
termestimatestd_errorstatisticp_values_valueconf_lowconf_high
strf64f64f64f64f64f64f64
"b2=b1"0.0649140.0165383.9250540.00008713.4934390.0324990.097328
" ], "text/plain": [ "shape: (1, 8)\n", "┌───────┬──────────┬───────────┬──────┬──────────┬──────┬────────┬────────┐\n", "│ Term ┆ Estimate ┆ Std.Error ┆ z ┆ P(>|z|) ┆ S ┆ 2.5% ┆ 97.5% │\n", "│ --- ┆ --- ┆ --- ┆ --- ┆ --- ┆ --- ┆ --- ┆ --- │\n", "│ str ┆ str ┆ str ┆ str ┆ str ┆ str ┆ str ┆ str │\n", "╞═══════╪══════════╪═══════════╪══════╪══════════╪══════╪════════╪════════╡\n", "│ b2=b1 ┆ 0.0649 ┆ 0.0165 ┆ 3.93 ┆ 8.67e-05 ┆ 13.5 ┆ 0.0325 ┆ 0.0973 │\n", "└───────┴──────────┴───────────┴──────┴──────────┴──────┴────────┴────────┘\n", "\n", "Columns: term, estimate, std_error, statistic, p_value, s_value, conf_low, conf_high" ] }, "execution_count": 38, "metadata": {}, "output_type": "execute_result" } ], "source": [ "# Yes!\n", "me.slopes(model5,\n", " newdata=slopes_datagrid,\n", " variables='yearsmarried', # average over this\n", " by='education', # for each of these\n", " hypothesis='b2 = b1')" ] }, { "cell_type": "markdown", "id": "bd536fd4-b321-418a-a662-57f4376ddc0c", "metadata": {}, "source": [ "### Answering bespoke questions\n", "You have seen so far how the model-based approach can answer some more complex questions than are usually posed and answered by psychology. We can now turn to an advanced use case, which is to use models to answer questions about individuals.\n", "\n", "Imagine a female friend who has been married to her partner for 10 years, has a high level of educational attainment, and is aged 32. Her partner and her are considering having a child. She wants to know how this will change her marital happiness. How can we help? Traditional statistics in psychology cannot, despite this being a very psychologically based question. Our model can answer this for us. All we need do is build a model that considers the interactions of these variables, and have it predict two things.\n", "\n", "The first should be our friends characteristics exactly as they are now. \n", "The second should be our friends altered characteristics, e.g., she then has a child.\n", "\n", "Then we can work out the difference and provide an answer. Lets do this, using the exact same logic we have already seen.\n", "\n", "First, the model." ] }, { "cell_type": "code", "execution_count": 43, "id": "97faa05d-b18c-4ee2-84e3-3b54264a1c84", "metadata": {}, "outputs": [ { "data": { "text/html": [ "
\n", "shape: (2, 10)
genderyearsmarriedageeducationchildrenrownamesaffairsreligiousnessoccupationrating
stri64i64i64stri64i64i64i64i64
"female"103220"no"1720455
"female"103220"yes"1720455
" ], "text/plain": [ "shape: (2, 10)\n", "┌────────┬──────────────┬─────┬───────────┬───┬─────────┬───────────────┬────────────┬────────┐\n", "│ gender ┆ yearsmarried ┆ age ┆ education ┆ … ┆ affairs ┆ religiousness ┆ occupation ┆ rating │\n", "│ --- ┆ --- ┆ --- ┆ --- ┆ ┆ --- ┆ --- ┆ --- ┆ --- │\n", "│ str ┆ i64 ┆ i64 ┆ i64 ┆ ┆ i64 ┆ i64 ┆ i64 ┆ i64 │\n", "╞════════╪══════════════╪═════╪═══════════╪═══╪═════════╪═══════════════╪════════════╪════════╡\n", "│ female ┆ 10 ┆ 32 ┆ 20 ┆ … ┆ 0 ┆ 4 ┆ 5 ┆ 5 │\n", "│ female ┆ 10 ┆ 32 ┆ 20 ┆ … ┆ 0 ┆ 4 ┆ 5 ┆ 5 │\n", "└────────┴──────────────┴─────┴───────────┴───┴─────────┴───────────────┴────────────┴────────┘" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "# Fit the model with all variables in it\n", "individual_mod = smf.ols('rating ~ gender * scale(yearsmarried) * scale(age) * scale(education) * children', data=affairs).fit()\n", "\n", "# Create a datagrid that fits the profile\n", "profile = me.datagrid(individual_mod,\n", " gender='female',\n", " yearsmarried=10,\n", " age=32,\n", " education=20,\n", " children=['no', 'yes']\n", " )\n", "\n", "# Show the profile\n", "display(profile)" ] }, { "cell_type": "markdown", "id": "086d0f16-7187-45ae-8ba5-cef685d89b83", "metadata": {}, "source": [ "All we now need do is recover these predictions and examine the difference." ] }, { "cell_type": "code", "execution_count": 44, "id": "8341ae5e-69ab-4d9c-bac1-ad1468c69f90", "metadata": {}, "outputs": [ { "data": { "text/html": [ "
\n", "shape: (2, 18)
genderyearsmarriedageeducationchildrenrowidestimatestd_errorstatisticp_values_valueconf_lowconf_highrownamesaffairsreligiousnessoccupationrating
stri64i64i64stri32f64f64f64f64f64f64f64i64i64i64i64i64
"female"103220"no"03.2334790.7266114.4500810.00000916.8299551.8093474.657611720455
"female"103220"yes"14.4147540.28476115.5033910.0inf3.8566344.9728751720455
" ], "text/plain": [ "shape: (2, 12)\n", "┌────────┬──────────────┬─────┬───────────┬───┬──────────┬──────┬──────┬───────┐\n", "│ gender ┆ yearsmarried ┆ age ┆ education ┆ … ┆ P(>|z|) ┆ S ┆ 2.5% ┆ 97.5% │\n", "│ --- ┆ --- ┆ --- ┆ --- ┆ ┆ --- ┆ --- ┆ --- ┆ --- │\n", "│ str ┆ str ┆ str ┆ str ┆ ┆ str ┆ str ┆ str ┆ str │\n", "╞════════╪══════════════╪═════╪═══════════╪═══╪══════════╪══════╪══════╪═══════╡\n", "│ female ┆ 10 ┆ 32 ┆ 20 ┆ … ┆ 8.58e-06 ┆ 16.8 ┆ 1.81 ┆ 4.66 │\n", "│ female ┆ 10 ┆ 32 ┆ 20 ┆ … ┆ 0 ┆ inf ┆ 3.86 ┆ 4.97 │\n", "└────────┴──────────────┴─────┴───────────┴───┴──────────┴──────┴──────┴───────┘\n", "\n", "Columns: gender, yearsmarried, age, education, children, rowid, estimate, std_error, statistic, p_value, s_value, conf_low, conf_high, rownames, affairs, religiousness, occupation, rating" ] }, "execution_count": 44, "metadata": {}, "output_type": "execute_result" } ], "source": [ "# Make the predictions\n", "me.predictions(individual_mod, newdata=profile)" ] }, { "cell_type": "markdown", "id": "c043e631-9e33-4df2-8a23-cb3ff39c1a13", "metadata": {}, "source": [ "And the contrast:" ] }, { "cell_type": "code", "execution_count": 49, "id": "e11958b6-ca07-42e8-b525-5a99b4e2161c", "metadata": {}, "outputs": [ { "data": { "text/html": [ "
\n", "shape: (1, 8)
termestimatestd_errorstatisticp_values_valueconf_lowconf_high
strf64f64f64f64f64f64f64
"b2=b1"1.1812750.7804181.5136440.1301162.942129-0.3483162.710867
" ], "text/plain": [ "shape: (1, 8)\n", "┌───────┬──────────┬───────────┬──────┬─────────┬──────┬────────┬───────┐\n", "│ Term ┆ Estimate ┆ Std.Error ┆ z ┆ P(>|z|) ┆ S ┆ 2.5% ┆ 97.5% │\n", "│ --- ┆ --- ┆ --- ┆ --- ┆ --- ┆ --- ┆ --- ┆ --- │\n", "│ str ┆ str ┆ str ┆ str ┆ str ┆ str ┆ str ┆ str │\n", "╞═══════╪══════════╪═══════════╪══════╪═════════╪══════╪════════╪═══════╡\n", "│ b2=b1 ┆ 1.18 ┆ 0.78 ┆ 1.51 ┆ 0.13 ┆ 2.94 ┆ -0.348 ┆ 2.71 │\n", "└───────┴──────────┴───────────┴──────┴─────────┴──────┴────────┴───────┘\n", "\n", "Columns: term, estimate, std_error, statistic, p_value, s_value, conf_low, conf_high" ] }, "execution_count": 49, "metadata": {}, "output_type": "execute_result" } ], "source": [ "me.predictions(individual_mod, newdata=profile, hypothesis='b2=b1')" ] }, { "cell_type": "markdown", "id": "252bacfa-6c29-4e0e-bc77-1256c0452dd4", "metadata": {}, "source": [ "According to the model, while our friend would see a rather large jump in their satisfaction of 1.18 units, this is a very uncertain outcome, and is not statistically significant. " ] } ], "metadata": { "kernelspec": { "display_name": "Python 3 (ipykernel)", "language": "python", "name": "python3" }, "language_info": { "codemirror_mode": { "name": "ipython", "version": 3 }, "file_extension": ".py", "mimetype": "text/x-python", "name": "python", "nbconvert_exporter": "python", "pygments_lexer": "ipython3", "version": "3.11.8" } }, "nbformat": 4, "nbformat_minor": 5 }