Variable __000000 already defined error (revisited) / eststo esttab error / "type mismatch error"

[Castor Comploj]

I am not sure if this is a bug, but I would consider it one (or two)

I define locals at the top of my .do file, and generate variables or run regressions based on those locals.
However, something inside Stata seems to have changed, but the problem already happens right after loading the dataset.

log     close _all
    macro drop _all
    cap drop __000000
clear     all
eststo  clear

I run the following code

        reg cesd10a4std i.q, robust
        eststo diff1
        *esttab diff*
        macro list
        cap drop __000000
        esttab diff*

If I do not run the line cap drop __000000, the line -esttab diff*- produces the the error "type mismatch".
However, macro list that diff1 is stored in eststo. -cap drop __000000- removes this eststo, and then the code runs again.

.                 macro list
eststo_counter: 1
eststo:         diff1
S_E_depv:       cesd10a4std
S_E_cmd:        regress
S_E_vce:        Robust
outpath:        G:\Shared drives\sharedStefan\---CHARLS\1-NRPS\outFiles
myFolder:       C:\Users\User\Documents\RUG\---CHARLS
myGoogle:       G:\Shared drives\sharedStefan\---CHARLS
S_level:        95
S_ADO:          BASE;SITE;.;PERSONAL;PLUS;OLDPLACE
S_StataSE:      SE
S_OS:           Windows
S_OSDTL:        64-bit
S_MACH:         PC (64-bit x86-64)
_rolloutyr:     rolloutyr
S_FN:           C:\Users\User\Documents\RUG\---CHARLS\CHARLSdata\H_CHARLS_D_DataPanelreg.dta
S_FNDATE:        3 Jul 2023 18:30

This solution seems very patchy though. What is the root cause of this issue? I tried reinstalling the package estout, but that did not solve the issue.
I do not think it is a Stata installation issue, since other files run as before. The problem must have to do with esttab or local macros.


Edit: The problem is solved throughout the whole file when I put -cap drop __000000- right after my global directory paths, but I do not use any globals after this, only locals.

[Nils Enevoldsen]

This seems difficult to diagnose without seeing more of your code and data. Please read Statalist FAQ 12. Remove as much as you can from your do-file and data while still being able to reproduce the "type mismatch" error, then post them both here.

[Castor Comploj]

I seem to have found the solution to this (and its root cause). I confirm this is a bug of either esttab, or tempvar.
In the data preparation file, I run the following:

    tempvar test
    gen     `test'=(cesd10a==. & cesd10a4!=.)
    reg     `test' age male ruralrsp wave if wave==1
    tempvar other 
    gen `other' =0

This stores the following variables, locally (not globally):

_other:         __000001
_test:          __000000

However, when moving to the next file and I run a regression, I use -eststo m1-. When using -esttab m1-, I get the error "type mismatch". At the same time, I noticed that inside -estpost-, I get the error variable __000000 already defined. This problem is also solved if I do not use tempvar in the data preparation do file.

See the trace of the last error using -qui estpost sum `mainvars' `othervars' `w' if `sample2' & `grp'==1 `estpost_add'-

      = version 17: estpost_summarize  cesd10a4 cesd10a4std pubpen ipubpen age iwyr male cnear higov nchild  if s_twfe & nrps2011==1
>  & s_ddr==1
        ---------------------------------------------------------------------------------------- begin estpost.estpost_summarize ---
        - version 8.2
        - local caller : di _caller()
        - syntax [varlist] [if] [in] [aw fw iw] [, ESample Quietly LISTwise CASEwise Detail MEANonly ]
        - if "`casewise'"!="" local listwise listwise
        = if ""!="" local listwise listwise
        - if "`listwise'"!="" marksample touse
        = if ""!="" marksample touse
        - else {
        - marksample touse, nov
variable __000000 already defined
          _estpost_markout2 `touse' `varlist'
          }
        ------------------------------------------------------------------------------------------ end estpost.estpost_summarize ---
      -------------------------------------------------------------------------------------------------------------- end estpost ---
      }
    --------------------------------------------------------------------------------------------------- end eststo._eststo_store ---
    exit
    }

I consider this a bug.

[Nils Enevoldsen]

Ah, this is https://www.statalist.org/forums/for...ng-in-data-set

It's unrelated to esttab. The behavior is set up by tempvar and save, and in this case is triggered by marksample (which appears to tempvar in a trenchcoat). Looks like this behavior goes back to at least Stata 14, possibly earlier. I like the suggestion to drop __00* before saving. It sure seems like unintended behavior, but I'll report to StataCorp and report back if they give me more info by email.

[Nils Enevoldsen]

I got a response and posted about it here: https://www.statalist.org/forums/for...-with-datasets

The short answer is you should drop all your temporary variables in your "data preparation do file" prior to saving.