unknown egen function group() When using fitstat

[Martin Orr]

I am currently running an ordered probit model and I want to look at the scalar measured of fit using fitstat, however I get this error message when I run the command post regression unknown egen function group(). If anyone can offer some advice as to why this is happening it would be highly appreciated.

Thanks

[Nick Cox]

The egen function group() has been in Stata a long time.

You need to tell us more about your exact code and ideally reproduce the error with data you give us or which are accessible to all.

Trying an ordered probit model on the rep78 variable from the auto dataset might be an example.

fitstat is a community-contributed command, so you are asked to say where you downloaded it from. Further, showing the results of

which fitstat

would do no harm.

[Martin Orr]

. which fitstat
c:\ado\plus\f\fitstat.ado
*! version 1.6.4 2/22/01 add warning messages

This is what comes up when I type which fitstat

[Nick Cox]

That's one piece of information. It doesn't let me guess at why this happens, but my advice is to upgrade.

I can find fitstat

*! 4.1.10 | 2017-01-05 | freese long | fixed bug in tobit (for v <14 and 15+)

and that is 16 years on from your version.

The only reason I can think of for not upgrading is that you are using a very early version of Stata. See https://www.statalist.org/forums/help#version for the advice of telling us about that.

[jill urbaeva]

I am using Stat 15, fairly new as far as I know.

[Nick Cox]

So, are jill urbaeva and Martin Orr the same person?

[jill urbaeva]

No, I am just Jill. I have no idea who Martin Orr is.

[Nick Cox]

OK; so you are just replying in the wrong thread. Not a problem.

[jill urbaeva]

Ah I see. Sorry, going back to my orignial thread.

[Richard Williams]

The current version of fitstat is part of Long and Freese's spost13 package. Uninstall the version you have. Then type

findit spost13_ado

and install it. If problems persist let us know.

If, by some chance, you are using Stata 9 or something like that, instead do

findit spost9_ado

Or, if your Stata is even more ancient,

findit spostado

The version of fitstat that is on SSC does more harm than good. Every few months we get a post like this, and every few months we have to tell people to get the right version. I will suggest to Scott Long that he delete or rename the SSC version.

[Nick Cox]

The version of fitstat on SSC uses the function in question just once at

            egen `depvgrp' = group(`depv') if e(sample)==1

Can anyone more familiar with fitstat than I am (very easy) or with egen (a bit harder) explain why that would ever fail?

[Martin Orr]

The current version of fitstat is part of Long and Freese's spost13 package. Uninstall the version you have. Then type

findit spost13_ado

and install it. If problems persist let us know.

If, by some chance, you are using Stata 9 or something like that, instead do

findit spost9_ado

Or, if your Stata is even more ancient,

findit spostado

The version of fitstat that is on SSC does more harm than good. Every few months we get a post like this, and every few months we have to tell people to get the right version. I will suggest to Scott Long that he delete or rename the SSC version.

This worked the fitstat command works now thanks!