Margins after Streg

[Alex Lampert]

Hello,

I am estimating marginal effects on discrete data after running streg. When I use predict(median time) as an option in my margins command, the estimates are not discrete nor do they end with 0.5. Am I misunderstanding the option? Does this option mean that I am estimating the effects by plugging in the median values of the covariates of the model into the coefficients?

Thanks!

[Marcos Almeida]

According to the Stata Manual;


"Options for predict:
median time calculates the predicted median survival time in analysis-time units. This is the prediction from time 0 conditional on constant covariates".

[Alex Lampert]

Thanks Marcos. I'm not sure I understand what that means. Does that mean it takes the predictions of all observations and then uses the median value of those predictions? For example, if I had 3 observations that created the following predictions: (3.1, 4.4, 5.1) my prediction would be 4.4?

[Marcos Almeida]

Please present some data - a toy example or an abridged version of the "real" deal may do the trick.

If you still think the information from the Stata Manual was not clear enough, I believe sharing data will be the best way to clarify your doubt overall.