Ok, I’ve sorted it out. It’s not a bug after all, but there is a typo in the documentation. Everything works if the restriction is (female = 0), but not with (female == 0). The documentation specifies “==” in the “var2 op val” line & below. I didn’t know this feature existed before today – I think my students will really like it! Cheers, Peter From: Sven Schreiber <sven.schreiber(a)fu-berlin.de&gt; Sent: Tuesday, September 3, 2024 11:42 AM To: gretl-users(a)gretlml.univpm.it Subject: [Gretl-users] Re: [EXTERNAL] Re: bug in test stat calculator Am 03.09.2024 um 17:36 schrieb Summers, Peter: Hi Sven, Yes, I typed the restrictions in parentheses and hit the enter key as described. Is the latest snapshot more recent than 08-05? Actually I don't expect that the vintage of the snapshot makes a difference in this case (although you never know...). Could you try with a public/example dataset that comes with gretl and be very precise which restrictions you are trying? Perhaps even a small screenshot. Maybe you're trying something that leaves one subsample empty, or similar. cheers sven