Can't cluster when using pantob（Tobit and fixed effects)

[Smith Shao]

Hello everyone!
I am using an unbalanced panel data from 2011 to 2022 to study the impact of the public search index of listed companies on OFDI. I hope to use pantob's tobit model to control the industry, province, and year fixed effects and cluster at the firm level. My other regression is mmqreg, but I encountered the problem of being unable to cluster when using pantob. I also tried xtpoisson regression but I still encountered the problem of being unable to cluster. Please tell me what I should do, or is there any other better way?
Thanks in advance!

[Carlo Lazzaro]

Smith Shao:
come comments about your post:
1) as per FAQ, please use CODE delimiters instead of screenshots to share what you coded and what Stata gave you back. Thanks;
2) the -pantob- routine is nit an official Stata module, As per FAQ, you're kindly requested to report this type of commands as community-contributed module. Thanks;
3) as -xtpoisson- does not allow clustered standard errors, probably the community-contributed module -pantob- does not allow it either. Go bootstrap standard error, if feasible.

[Smith Shao]

Hi Carlo Lazzaro:
Thanks for your kind advice.Here is the code:

pantob lnto esg_index age1 nstaff soe roa lev capital dyear2-dyear12 prov2-prov31 incode2-incode257

 xtpoisson lnto lnindex age1 nstaff soe roa lev capital i.year i.provid,fe

What do you mean by this sentence "As per FAQ, you're kindly requested to report this type of commands as community-contributed module"?I guess you mean I should introduce pantob.
So pantob is a model used to do tobit with fixed effects ,and it's available from https://www.princeton.edu/~honore/st...ob_version_0.6.
Regarding your third suggestion, do you think the poisson regression on my case is feasible as ofdi is not a count variable?In addition,I've run -tobit- with lsdv, and I'm worried that the results are inconsistent.

[Carlo Lazzaro]

Smith Shao:
1) please post also what Stata gave you back within CODE dlimiters. Thanks;
2) as per FAQ you're kindly requested to define non-official Stata commands as community-contributed modules (actually, you downloaded -pantob- from outside the Stata library). The reason underlying this recommedantion rests on the evidence that the same community-contributed module may have different releases floating around the cyberspace;
3) -tobit- and -poisson- regressions have different theoretical roots. That sais, -poisson- with -vce(cluster clusterid)- is an option. The issue on how to retrieve the (conditional) fixed effect you're interested in still remains, though.

[Smith Shao]

1.Pantob (version 0.6).

pantob implements the estimators developed in Honoré (1992).
Date: 8/27/2009; slight revision 9/15/2009

https://www.princeton.edu/~honore/stata/

2.Firstly I tried -pantob- but cluster wasn't allowed and the code of -xtpoisson- is similar.

pantob lnto esg_index age1 nstaff soe roa lev capital dyear2-dyear12  incode2-incode257,cluster(inid)

option cluster() not allowed

 xtpoisson lnto lnindex age1 nstaff soe roa lev capital i.year i.provid,fe vce(cluster inid)

option cluster not allowed

so I tried -tobit- and it worked out, but I'm not sure whether it's appropriate

 tobit lnto lnindex age1 nstaff soe roa lev capital dyear2-dyear12  incode2-incode257,ll(0) vce(cluster code) nolog

Tobit regression                                    Number of obs     = 11,016
                                                           Uncensored = 11,016
Limits: Lower =    0                                    Left-censored =      0
        Upper = +inf                                   Right-censored =      0

                                                    F(202, 10815)     =      .
                                                    Prob > F          =      .
Log pseudolikelihood = -23601.887                   Pseudo R2         = 0.0523

                               (Std. err. adjusted for 2,541 clusters in code)
------------------------------------------------------------------------------
             |               Robust
        lnto | Coefficient  std. err.      t    P>|t|     [95% conf. interval]
-------------+----------------------------------------------------------------
     lnindex |      0.182      0.044     4.17   0.000        0.096       0.267
        age1 |     -0.010      0.008    -1.23   0.218       -0.025       0.006
      nstaff |      0.000      0.000    10.82   0.000        0.000       0.000
         soe |      0.076      0.095     0.80   0.426       -0.111       0.263
         roa |      0.528      0.059     8.97   0.000        0.413       0.644
         lev |     -1.316      0.206    -6.40   0.000       -1.719      -0.913
     capital |      0.023      0.016     1.43   0.154       -0.009       0.056
      dyear2 |      0.296      0.765     0.39   0.699       -1.203       1.796
      dyear3 |      0.602      0.768     0.78   0.433       -0.904       2.10

For the sake of brevity I only show part of the results.

Can you give me further guidance?
Thanks in advance!