Comparing 2 model predictions - how do I do this?

[Jake Wu]

Hi all,

I've developed a multiple logistic regression model with ~50 predictors (and their associated coefficients). These predictors encompass various demographic profiles (income, gender, age)

I want to get some insights by comparing specific scenarios (e.g. low income males vs high income males, older females vs younger females, etc. etc.)

In the interest of avoiding a hand-calculation - would anyone be able to point me towards specific features of STATA that allow me to do such scenario comparisons - and provide associated comparison data e.g. standard errors, significance levels if any, etc?

Appreciate your guidance and apologies in advance if I'm missing out anything glaring.

Thanks again

[Andrew Musau]

I want to get some insights by comparing specific scenarios (e.g. low income males vs high income males, older females vs younger females, etc. etc.)

Assuming that these are levels of a categorical variable, see

help test

[Leonardo Guizzetti]

See also -margins- which I think is more directly what you are after, and also more relevant for prediction modeling in general. Margins will (by default) give you estimated marginal means or their differences. It will also by default give you statistical significance of these estimates and I would caution you not to pay (much) attention to those as you may inadvertently test hypothesis of no scientific interest. Proper evaluation of prediction models will always require validated using a new, independent dataset.