[George Ford]

That didn't work either. Still different. But only different in the percentile results. First set is identical.

runiform() accepts arguments in v18. The help file says it does.

[daniel klein]

runiform() accepts arguments in v18. The help file says it does.

Indeed; my mistake. Appearently, runiformint() is not implemented as floor(runiform()).

[George Ford]

The replacement may be fine and perhaps I should not expect the results to be exactly the same even with the same seed. If I generate random random numbers using the same seed between the two approaches I do not get the same results (but the same results for the same function run twice, resetting the seed, as expected). I suppose the results may be different, but simply due to the differences in the way the seed works between the two.

Changing the seed in the stock approach produces a range of results that encompass the results of the alternate approach.

[George Ford]

I ran the stock code 20 times to get a range of the reported statistics. The alternative approach is pretty typical of the results. So I think the alternative works.

[Evamgelos Salachas]

Hi all,

Thank you for the suggestions. Unfortunately, nothing worked in the Stata 13 version I use.

However, I manage to find a Stata 18 from a colleague and I run again the code. However, this time the following message appears:

tvgc_cv(): 3499 mm_quantile() not found
<istmt>: - function returned error

[Leonardo Guizzetti]

For the new error you need to install moremata from SSC.

[Nick Cox]

Evamgelos Salachas If your name is really Evangelos S, you can get in touch with sladmin (Statalist administrator) to fix it. Use the CONTACT US button at bottom right. Naturally this is your call, especially if that is your real name.

[Evamgelos Salachas]

Hopefully, I installed moremata in Stata 18 and I manage to run the code.

Thank you very much for the support.