Re: st: Using suest after mfx2 or margeff - error r(322)

[Amanda Fu <mandy.fu1@gmail.com>]

Hi all,

I would like to test the difference of the marginal effects (not
coefficients) for a variable across probit regressions. When checking
previous discussion, I find my question is exactly the same with the
following early one, but no one answered it.

I think there must be something wrong with my question. Would someone
please give me some hints? Thanks!
---------------------------------
probit y a1 a2 a3 if male==1
mfx compute
est store MALE

probit y a1 a2 a3 if female==1
mfx compute
est store FEMALE
suest MALE FEMALE
test [MALE_mean]a1=[FEMALE_mean]a1
----------------------------------------
Amanda

On Sun, Nov 21, 2010 at 9:10 AM,  <pascalstock@freenet.de> wrote:
> Hello STATA users,
>
> I try to use the suest command after estimating two different sets of marginal effects at the mean with mfx2 or margeff from two different probit regressions. I intend to compare the differences between the marginal effects betwenn the two probit regressions whether they differ significantly with Wald tests. For this I first want to use suest to have one set of marginal effects from both probit regressions. Unfortunately the stored marginal effects, as if mfx2 or margeff were estimation commands, provide the following error message
>
>
> .  suest unadvised non_bulge_bracket
> unable to generate scores for model unadvised
> suest requires that predict allow the score option
> r(322);
>
> "unadvised" and "non_bulge_bracket" are the names I gave to the stored marginal effects:
>
> probit follow_on_bid industry_experience_deals_A ... if advisory_choice==1
> margeff, at(mean) replace
> estimates store unadvised
>
> probit follow_on_bid industry_experience_deals_A ... if advisory_choice==2
> margeff, at(mean) replace
> estimates store non_bulge_bracket
>
> suest unadvised non_bulge_bracket
>
> test [unadvised=non_bulge_bracket], common
>
> test [unadvised]industry_experience_deals_A=[non_bulge_bracket]industry_experience_deals_A
> . ..
>
> With regular probit regression coefficients it works perfectly, but not with the stored marginal effects. Do you have an idea how I can proceed?
>
> Kind regards and thanks for your help
>
> Pascal
>
>
>
>
> freenetMobile - Handytarif des Jahres!
> Für 8 Cent/Min. telefonieren & kostenlos mobil surfen!
> Sichern Sie sich jetzt das Startpaket mit 10 &euro; Startgutgaben unter http://www.freenetmobile.de/topangebot/?pid=10111947014!
>
>
> *
> *   For searches and help try:
> *   http://www.stata.com/help.cgi?search
> *   http://www.stata.com/support/statalist/faq
> *   http://www.ats.ucla.edu/stat/stata/

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/