[Alejandro Torres]

Dear Carlo,

once again, than you very much for taking the time and for the clear answer. I really appreciate.
I`ll try not caoming back with more questions (soon).
Thank you so much again,

Alejandro

[Carlo Lazzaro]

Alejandro:
you'll be welcomed whenever you'll decide to post again on this forum.
Luckilly enough, this forum does not require any sort of points card that gets emply after a given number of posted topics!
Hence, feel free to post whenever you feel you can benefit from this forum, the only requirement being acting on the FAQ (like you did, so far).

[Alejandro Torres]

Dear Carlo, I am sorry but I need to go back to you again please.

I just did what you recommended me:

tobit DV IV i.year i.industry i.firm, vce(cluster firm)

but, I received the following message:

maxvar too small
    You have attempted to use an interaction with too many levels or attempted to fit a model with too many variables.  You need to
    increase maxvar; it is currently 5000.  Use set maxvar; see help maxvar.

    If you are using factor variables and included an interaction that has lots of missing cells, either increase maxvar or set
    emptycells drop to reduce the required matrix size; see help set emptycells.

    If you are using factor variables, you might have accidentally treated a continuous variable as a categorical, resulting in lots
    of categories.  Use the c. operator on such variables.

(I am using stata 15)

Now, I am thinking that I have more that 35,000 firms, so, may be there is to many dummies to create, could be that the reason?

Then I tried the same, but without the i.firm

tobit DV IV i.year i.industry , vce(cluster firm)

That worked, but it is correct doing the vce (cluster firm) with out the i.firm?

Thank you so much again,

Alejandro

P.S. Thank you for your kind message before.

[Carlo Lazzaro]

Alejandro:
1) the first warning message that you received is actually due to your asking too much out of your copy of Stata in terms of variables creation.
2) it's absolutely fine clustering on firm without -i.firm-. Obviously, this is a different regression model and cannot give you back fixed effect.
More substantively, I think you can be interested in this thread: https://www.statalist.org/forums/for...-fixed-effects.

[Alejandro Torres]

Thank you very much Carlo !! All is starting to make more sense now, thank you for the link also, what a great "conversation/post".
Once again, thank you so much for taking the time.

Alejandro

[Jayati Chatterjee]

Dear Carlo,
I have a doubt regrading clustering, I am using pooled oprobit model, and when I am clustering at industry(sector) or state level with my industry as my control the Wald chi2 is missing, if I drop the industry control the model seems to be working fine.
I read it up on help j_robustsingular but bootstrapping or jackknife does not seem to be working.
I don't think I can cluster on firmid level. Could you suggest something?

[Carlo Lazzaro]

Jayati:
without further details from your side (as reported in the FAQ, phrasing like "it does not work" are not that informative ), the -vce- options available for -oprobit- are: oim, opg, robust, cluster clustvar, bootstrap, or jackknife.
On a different note, clustering makes sense with at least 30 -clusterid-.

[Jeff Wooldridge]

Jayati: I suspect you have many industries, which is likely causing problems in the oprobit when you include the industry fixed effects. Including lots of FEs in nonlinear models usually causes an incidental parameters problem. As Carlo points out, it would be helpful to provide many more details -- including showing us your data and Stata code.