Old bug

[Sergiy Radyakin]

Dear All,

I hope someone following Stata for many years can help me with remembering: there used to be a problem (and I hope I can recollect correctly) where a method was introduced and then later withdrawn by the StataCorp. Here is what I remember, any or all of the below may be wrong:

- it was related to calculation of standard errors or confidence intervals;
- with survey data
- probably introduced in version 8 and recalled (disabled) in version 9, or in the 9 to 10 transition, but unlikely anything newer.
- could be survey proportion or survey ratio commands, or something similar.
- could be https://www.stata.com/statalist/arch.../msg00126.html , but I somehow remember this regarding the SVY subset of commands.

I believe this was mentioned earlier in the Statalist, but googling now didn't result in any mentioning of that problem.

If anyone can point to more details, please do.

Much appreciated, Sergiy Radyakin

[William Lisowski]

Have you tried searching for svy in the Viewer window for

help whatsnew7to8
help whatsnew8
help whatsnew8to9
help whatsnew9
help whatsnew9to10

[Sergiy Radyakin]

Thank you, yes, I tried, but since the word "error" is mentioned almost in every line (in the context of "standard errors" typically) that was not very fruitful. Plus the language is sometimes "xyz was reported inappropriately" or "inadequate" or something else, so I am not even sure how it was formulated. I checked with the most recent version, and I believe for heads on proportion or ratio the errors are reported now. Since I was part of this discussion, I believe that it may have been that case, but I am recollecting someone from StataCorp posted later explaining what went wrong and how it was corrected, and I can't find that post.