Probit interaction effects

The task

Task statement

Fit a Bernoulli probit model by maximum likelihood on all rows, with intercept and regressors x, d, x*d, z, and x squared. Do not weight observations. For the subset target==1, calculate the average marginal effect of continuous x, including both its interaction and squared term, and report ame_x. Also calculate the average discrete effect of changing d from 0 to 1 in this same subset, holding x and z at each row's observed values and updating x*d consistently, and report effect_d. Fit the model once on the full sample. Average individual effects over the stated target subset. Do not evaluate effects at mean covariates. For both rows, n_obs is the full estimation sample size. Evaluate the x derivative at each target row's observed x, d, and z.

Data

input.csv: y = Dependent outcome; x = Continuous regressor; d = Binary regressor; z = Additional exogenous covariate; target = Indicator for the population over which probability effects are averaged

Files

input.csv

Required output

Scored estimates: ame_x, effect_d.

Reference output

A run succeeds when every scored estimate lies within 1 percent or 0.005 of its reference value, whichever is larger. Sample sizes are not scored.