[Gretl-users] Re: 2021b git: WLS estimation bug

On Sat, 17 Apr 2021, Sven Schreiber wrote: > Am 17.04.2021 um 09:13 schrieb Artur Bala: >> Dear Allin, >> >> In a WLS estimation when the weight variable contains zeros gretl stops >> executing saying that "Weight variable is not a dummy but contains zeros" >> The same worked perfectly before with the observations weighted 0 being >> simply dropped >> from the estimation. >> > > Yes, that is an intentional change, because zero weights mess up the > definition/concept of the sample size and so on. If you want to exclude > certain observations, you now will have to explicitly restrict the > sample to achieve the same effect. Or, if your choice of zero was just a > shortcut to mean "really small", you can assign a "really small" but > strictly positive weight instead. > > To put it differently, we reckoned that the earlier behavior could have > been considered a (minor) bug. In fact, we discussed this issue back in January: thread here: https://gretlml.univpm.it/hyperkitty/list/gretl-users@gretlml.univpm.it/t... or, if you prefer old-style mail archives, click the link below: https://www.mail-archive.com/gretl-users@gretlml.univpm.it/msg15526.html