Joseph Coveney <[email protected]> is using a temporary variable with
-ml- in his ado-file estimation command but wants to show something other than
the tempvar's name in the output:
> Is there a way to provide an informative variable name in -ml display- when
> the ado-file has to make use of a temporary variable for the actual
> estimation?
>
> For example, after convergence using the ado-file pair below,
> -ml display- shows the response variable as __000000 in the regression
> table. I examined the documentation for -ml- and looked through W. Gould,
> J. Pitblado & W. Sribney, _Maximum Likelihood Estimation with Stata_ Third
> Edition. (College Station, Texas: Stata Press, 2006), but couldn't find any
> way to change the display to show the original dependent variable's name.
Joseph can use 'ereturn local' to change the contents of 'e(depvar)' after
-ml- fits the model. For example, Joseph can add the following line to
-firthlogit.ado- to make -e(depvar)- contain the variable name supplied by the
user:
ereturn local depvar `response'
I added the above line in Joseph's Stata code, which appears after my
signature.
--Jeff
[email protected]
> ---------------Main ado-file (-firthlogit.ado-)-------------
>
> *! firthlogit.ado Version 0.05 JRC 2008-02-11
> program define firthlogit, eclass
> version 10
> syntax varlist(numeric) [if] [in] [fweight] [, nolog]
> tempvar tmpvar0
> tempname tmpnam0
>
> marksample touse
>
> // Initial values
> gettoken response predictors : varlist
> generate byte `tmpvar0' = (`response' != 0) if `touse'
> summarize `tmpvar0' `weight' `exp', meanonly
> if inlist(r(mean), 0, 1) {
> display in smcl as text "outcome does not vary; remember:"
> display in smcl as text ///
> " 0 = negative outcome,"
> display in smcl as text ///
> " all other nonmissing values = positive outcome"
> error 99
> }
> scalar define `tmpnam0' = logit(r(mean))
>
> local k : word count `predictors'
> if (`k') matrix define `tmpnam0' = (J(1, `k', 0), `tmpnam0')
> else matrix define `tmpnam0' = (`tmpnam0')
>
> global firthlogitpredictors `predictors'
>
> ml model d0 _firthlogit ///
> (xb: `tmpvar0' = `predictors') if `touse' `weight' `exp', ///
> missing init(`tmpnam0', copy) maximize ///
> crittype("penalized log likelihood") `log'
>
> ereturn local depvar `response' // <- NEW LINE
> end
>
> ------------------------------------------------------------
>
> --------estimation ado-file (-_firthlogit.ado-)-------------
>
> program define _firthlogit
> version 10
> args todo b lnL
> tempvar xb mu se
> tempname I
>
> assert !`todo'
> mleval `xb' = `b', eq(1)
> quietly {
> generate double `mu' = invlogit( `xb') if $ML_y1 == 1
> replace `mu' = invlogit(-`xb') if $ML_y1 == 0
> mlsum `lnL' = ln(`mu')
>
> generate double `se' = sqrt(`mu' * (1 - `mu'))
> local mataccumlist
> foreach var of global firthlogitpredictors {
> tempvar `var'
> generate double ``var'' = `var' * `se'
> local mataccumlist `mataccumlist' ``var''
> }
> matrix accum `I' = `mataccumlist' `se', noconstant
> scalar define `lnL' = `lnL' + ln(det(`I')) / 2
>
> }
> end
>
> ------------------------------------------------------------
>
>
> *
> * For searches and help try:
> * http://www.stata.com/support/faqs/res/findit.html
> * http://www.stata.com/support/statalist/faq
> * http://www.ats.ucla.edu/stat/stata/
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
