Suest command: unable to generate scores error with different regression models

[Orifjon Kurbanov]

Dear all,

I have a following problem. Firstly, I am calculating 2 regressions for for two groups of subsample and testing the difference. Secondly, I am calculating 2 additional regressions for the other subsample groups and also testing the difference of coefficients. The task is to calculate the difference of previously calculated differences (sorry, if it sounds strange). At this stage Stata gives me an error unable to generate scores for model storedresults* suest requires that predict allow the score option.

I have a following code:

Stage #1
reg alpha MKTRF if status==1 & port==1
est sto capmst_long
reg alpha MKTRF if status==1 & port==2
est sto capmst_short
suest capmst_long capmst_short
lincomest [capmst_long_mean]_cons - [capmst_short_mean]_cons
est sto capm1

Stage #2
reg alpha MKTRF if status==2 & port==1
est sto capmnst_long
reg alpha MKTRF if status==2 & port==2
est sto capmnst_short
suest capmnst_long capmnst_short
lincomest [capmnst_long_mean]_cons - [capmnst_short_mean]_cons
est sto capm2

Stage #3 with error 
suest capmst_long capmnst_long
lincomest [capmst_long_mean]_cons - [capmnst_long_mean]_cons
est sto capm_long_diff
suest capmst_short capmnst_short
lincomest [capmst_short_mean]_cons - [capmnst_short_mean]_cons
est sto capm_short_diff
suest capm_long_diff capm_short_diff

Would be very grateful for any suggestions!

[Joro Kolev]

For what you have shown before the error occurred, I think it is enough information.

However then pick up the code where the error occurs, and copy and paste exactly what you typed at Stata, and exactly what Stata returned.

[Orifjon Kurbanov]

For what you have shown before the error occurred, I think it is enough information.

However then pick up the code where the error occurs, and copy and paste exactly what you typed at Stata, and exactly what Stata returned.

Joro, this is what I type in Stata and the error, which occurs afterwards:

. suest capm_long_diff capm_short_diff
unable to generate scores for model capm_long_diff
suest requires that predict allow the score option
r(322);

[Joro Kolev]

I think then you are overestimating what -suest- can do. I think you take -suest- to be a general tool, that can do "things jointly" after every step you make.

But it is not like this. After some particular commands, -suest- can join them.

In this particular case you seem to be trying a super-suest, that is, you seem to be trying to apply -suest- on a previous -suest- result.

Joro, this is what I type in Stata and the error, which occurs afterwards:

. suest capm_long_diff capm_short_diff
unable to generate scores for model capm_long_diff
suest requires that predict allow the score option
r(322);

[Orifjon Kurbanov]

I think then you are overestimating what -suest- can do. I think you take -suest- to be a general tool, that can do "things jointly" after every step you make.

But it is not like this. After some particular commands, -suest- can join them.

In this particular case you seem to be trying a super-suest, that is, you seem to be trying to apply -suest- on a previous -suest- result.

Okay, thank you for this comment. I really might be thinking wrongly since I am not a proficient user of Stata yet.
Maybe you can advice some other solutions for my problem?

[Carlo Lazzaro]

Orif.
I'm probably late to the party (or have already had too much fun in a previous one ), but why not running something like:

regress alpha MKTRF i.status##i.port

and then -test- or -testparm-?

[Orifjon Kurbanov]

Orif.
I'm probably late to the party (or have already had too much fun in a previous one ), but why not running something like:

regress alpha MKTRF i.status##i.port

and then -test- or -testparm-?

Thanks, Carlo, I've tried this one, but I am not sure how to access the reference category if I want to test the difference between intercept of each group. Basically, I do not need coefficients other than _cons for each group.