Model Validation Using Stata

[Melody Brown]

Hi all,

I have a question about how to do model validation using Stata.

I used a training dataset to run regression and get the coefficients of the regressions. I want to first, save the result of these coefficients and second, use these coefficients in a new dataset to do prediction and calculate MSE, MPE, etc.

I have many variables and one categorical variable has around 3000 values. I wonder if anyone knows how to save the regression coefficients and then apply them to the new dataset to do prediction.

Thanks!

[Joseph Coveney]

How about something along these lines?

use training_dataset
generate byte training_dataset = 1
append using new_dataset
replace training_dataset = 0 if missing(training_dataset)
regress <stuff> if training_dataset
predict double outcome_hat, xb
<summary statistics of difference between observed and predicted outcome variables if !training_dataset>

[Jared Greathouse]

What's the point of doing this? Why are you interested in this?