last estimates not found

[Jerome Anderson]

I wrote this code:

oneway MEASURE Par1_Race, tabulate

| Summary of MEASURE
Par1_Race | Mean Std. dev. Freq.
------------+------------------------------------
Amer_Ind | 65.84627 22.986507 185
Asian | 63.754536 23.025959 1,120
Black | 61.720519 24.628945 1,619
Pac_Is | 67.693291 19.389287 79
White | 65.837978 21.551192 11,737
Mixed | 62.863707 23.545444 437
------------+------------------------------------
Total | 65.169119 22.116957 15,177

Analysis of variance
Source SS df MS F Prob > F
------------------------------------------------------------------------
Between groups 29657.2914 5 5931.45828 12.17 0.0000
Within groups 7393831.6 15171 487.366133
------------------------------------------------------------------------
Total 7423488.89 15176 489.159785

Bartlett's equal-variances test: chi2(5) = 65.9155 Prob>chi2 = 0.000

. pwcompare Par1_Race
last estimates not found

and, as you can see, the last estimates were not found. Why? I've run this code before, and it worked. Why would the last estimates not be found?

[Clyde Schechter]

The message "last estimates not found" is misleading. The problem is something only tangentially related to that. The problem is that -pwcompare- calls a program called _check_eclass. As _check_eclass is a built-in program its code and inner workings are not accessible to users. But _check_eclass returns a "last estimates not found" error message after -oneway-, whereas it does not do so with the more modern estimation commands, or even with -anova-. If you change -oneway- to -anova-, -pwcompare- will run and give you the results you seek.

[Jerome Anderson]

Clyde, thank you for this clarification. I did as you suggested, and the code ran. I could also do mcompare(tukey) pveffects successfully.

The one problem is anova does not allow tabulate. If you want the tabulate results, you must run oneway. Would be nice if tabulate worked w/ oneway.

Thanks again for your help. You are a wealth of info as to the inner workings of Stata.