Post estimation prediction without using some of the coefficients

[Pedro Santos]

My problem is how to make a post regression prediction by not using all the coefficients obtained in the regression. I would like to predict the outcome using only the year and age dummies. I solved the problem in a cumbersome way. I am wondering whether there is a more suitable command for that, as it would facilitate my calculations.

qui reg wage dum1-dum26 i.year i.age, r

I replace all the values of the coefficients dum1-dum26 with 0's. The command 'predict' delivers me the prediction using only the age and year dummies.

forv i = 1(1)26 {
     replace dum`i'=0 if dum`i'==1
}
predict yhat, xb

[Clyde Schechter]

I think that is the simplest way to accomplish it. I can think of at least one other approach, but it is more complicated.

[Pedro Santos]

Ok, thanks a lot Clyde!

[Phil Bromiley]

If you don't want to change your data, you could also just calculate the predicted value with a generate statement. After any command, issue command,coefl and Stata will tell you how to access the specific parameters.