st: Extended macro problem

[Fred Wolfe <fwolfe@arthritis-research.org>]

I hope a macro guru might be able to help me with this problem.

Either I have made a mistake or there is a Stata bug.

I am developing a program to format the output from Roger Newson's -parmby-

Here is a snippet of code.

program define showparmest
   version 8.2
   syntax using/ [, Clearfile Sort ciform(string)]
   preserve
   qui compress
   qui {

      // identify the type of command
      local command `e(cmd)'

      // identify categorical variables used with "i." in an xi
      local cmdline `r(command)'
      foreach word of local cmdline {
         if substr("`word'",1,2) == "i." {
            local tname = substr("`word'",3,.)
            local catlist `catlist' `tname'       // macro catlist 
comtains i.cat vars
         }                                        // with i. stripped off
      }

      // get levels and omitted category for catvars
      foreach catvar of local catlist {
         levels `catvar', local(`catvar'lev)                    // lev in 
VARNAMElev
         local `catvar'omit : char `catvar'[omit]            // omit in 
VARNAMEomit
         if "``catvar'omit'" == "" {
            local `catvar'omit : word 1 of ``catvar'lev'     // low cat if 
no char[omit]
         }


         foreach num of local `catvar'lev {                  // get value 
labels
            local `catvar'lab`num': label `catvar' `num'
            noi di "``catvar'lab`num''"
         }


      }
      noi des sex edcat
      noi lab list sex edcode2
      noi tabl edcat

The key section to pay attention to is the section beginning   // get value 
labels.
This section uses the extended macro function to assign value labels to macros.

In the //get value labels section, in the foreach cycle, -sex- is processed 
first then -edcat- . Both variables have value labels assigned but Stata 
only sees the labels for -sex-, not for -edcat-

Here is the output:

This is the parmby output to show you what I am doing. This is a nonsense 
regression for test purposes.

. parmby "xi:logistic married haq pain i.sex fatigue i.edcat age edlevel ", 
label eform format(estimate min95 max95 %8.2f p %8.1e) saving(killme,replace)
Command: xi:logistic married haq pain i.sex fatigue i.edcat age edlevel
i.sex             _Isex_0-1           (naturally coded; _Isex_0 omitted)
i.edcat           _Iedcat_0-4         (naturally coded; _Iedcat_2 omitted)

Logistic regression                               Number of obs   =       7415
                                                  LR chi2(10)     =     369.84
                                                  Prob > chi2     =     0.0000
Log likelihood = -4290.4222                       Pseudo R2       =     0.0413

------------------------------------------------------------------------------
     married | Odds Ratio   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
haq_disabi~y |   .8444004   .0413379    -3.45   0.001     .7671453    .9294355
  pain_scale |   .9873622   .0132877    -0.95   0.345     .9616593    1.013752
     _Isex_1 |    2.88479   .2257692    13.54   0.000     2.474557    3.363031
fatigue_sc~e |   1.003774   .0121381     0.31   0.755     .9802633    1.027848
   _Iedcat_0 |   .4256095   .1216269    -2.99   0.003     .2430874    .7451783
   _Iedcat_1 |   .6159096   .0804384    -3.71   0.000     .4768142    .7955816
   _Iedcat_3 |   1.073439   .1213579     0.63   0.531     .8600911    1.339708
   _Iedcat_4 |   1.294813    .301287     1.11   0.267     .8206213    2.043013
         age |    .980985    .002152    -8.75   0.000     .9767761     .985212
     edlevel |   .9171411   .0447198    -1.77   0.076     .8335498    1.009115
------------------------------------------------------------------------------
file killme.dta saved

Here is the output from my program snippet.

. showparmest using killme,  s

Female
Male
0
1
2
3
4

Notice that the value labels are shown for sex (Female Male) but not for 
-edcat (shown as 0 1 2 3 4 ).

              storage  display     value
variable name   type   format      label      variable label
-------------------------------------------------------------------------------
sex             byte   %8.0g       sex        Patient's gender
edcat           byte   %9.0g       edcode2    Education category
sex:
           0 Female
           1 Male
edcode2:
           0 0-8
           1 8-11
           2 12
           3 13-15
           4 16 or >

edcat (Education category)     code  |   Freq    Percent
-------------------------------------+--------------------
                       0-8       0   |    255       2.81
                      8-11       1   |    748       8.25
                        12       2   |   3468      38.25
                     13-15       3   |   2270      25.04
                   16 or >       4   |   2325      25.65
-------------------------------------+--------------------
           <missing value>       .   |    139
-------------------------------------+--------------------
                             Total   |   9205     100.00

Notice that the value labels are shown in -describe-, -label list- and -tabl.
Thus the labels are there but the program is not processing them.

I hope someone may have a suggestion as to my error. Or is this a Stata bug.

Thanks,

Fred



Fred Wolfe
National Data Bank for Rheumatic Diseases
Wichita, Kansas
Tel (316) 263-2125     Fax (316) 263-0761
fwolfe@arthritis-research.org


*
*   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/