Error r(322) - for mixed-effects regression and quantile regression

[Nouschka Snijders]

Dear all,

I would like to run a Multilevel mixed-effects linear regression (command Stata: mixed). Next to that also a Quantile regression (command Stata: qreg).

For both regressions I obtain the error r(322):
- first error: mixed is not supported by suest
- second error: qTD was estimated with a nonstandard vce (iid)

My Stata commands look as folllows (sorry I cannot give you the commands as given in FAQ, but it is not possible for me to download anything on this computer):

mixed TE1 VOL SKEW KURT DIV PD || Market:
predict mixedTE1
est store mixedTE1
regress TE1 VOL SKEW KURT DIV PD
predict generalTE1
est store generalTE1
suest mixedTE1 generalTE1

regress TE1 VOL SKEW KURT DIV PD
predict hatTE1
est store hatTE1
regress TE2 VOL SKEW KURT DIV PD
predict hatTE2
est store hatTE2
qreg TE1 VOL SKEW KURT DIV PD, quantile(0.05)
predict qTD
est store qTD
suest hatTE1 hatTE2 qTD


How can I solve this error r(322)? What do I do wrong?

Thank you in advance

[Clyde Schechter]

The first error message means exactly what it says: -suest- does not support -mixed-.

The second is more obscure, but it means the same as the first: -suest- also does not support -qreg-.

You didn't do anything wrong, but applying -suest- to these models is simply not possible.

[Nouschka Snijders]

Clyde Schechter Thank you!! Do you know another command I can use to restore the estimates of the mixed regression?

[Clyde Schechter]

If all you need to do is restore the estimates of the mixed regression, it's just:

estimates restore mixedTE1

[Nouschka Snijders]

Clyde Schechter Sorry, I meant that I have to use the stored estimates to run t-tests... is there a way I can do this?

[Clyde Schechter]

I meant that I have to use the stored estimates to run t-tests

I'm not sure what you mean by this.

If you mean that you want to run t-tests contrasting coefficients, as one would normally do with the -test- or -lincom- or -nlcom- commands directly after a regression, once you have run -estimates restore mixedTE1- you can use those same -test- etc. command exactly in the same way you would have immediately after the regression. -estimates restore- restores Stata's state to the state it was in just after that regression finished executing. (Well, you may have to restore the sample information. See -help estimates esample- for details.)

If you have some other kind of t-tests in mind, please clarify.

[Nouschka Snijders]

Clyde Schechter I would like to compare betas from an OLS regressin with betas from a quantile regression.

regress TE1 VOL SKEW KURT DIV PD
predict hatTE1
est store hatTE1
qreg TE1 VOL SKEW KURT DIV PD, quantile(0.05)
predict qTE1
est store qTE1
suest hatTE1 qTE1

test hatTE1 == qTE1

Then, I obtain the error:
qTE1 was estimated with a nonstandard vce (iid)
r(322);

How can I solve this?