I've been able to replicate this in versions as early as 6.1.4. As Bas pointed out, SPSS doesn't seem to be "counting" the filtered cases or weighing any case improperly; the results of whatever statistical procedure you run after weighing the data seem to be the same if you're filtering the file or permanently selecting a subset of cases. My guess is that SPSS assess all cases in the active file -- even filtered ones -- when you apply a weight variable, though only unfiltered cases are ultimately used in subsequent procedures. When it sees zero or negative values, it returns the #3211 warning whether the offending cases are filtered or unfiltered. I would also guess that that the SPSS developers consciously chose not to override the warning when offending cases are filtered, since the weighting will remain in effect if and when the filter is removed. The user would probably want to know that the weighting variable has these characteristics in general, even if all active cases are positive values under a particular filtering scheme.

Nonetheless, the warning might seem confusing and unduly alarming in cases such as the one Bas and Susan describe. I've logged this issue with the SPSS developers so they can take a look at it.

Ed Tverdek SPSS Technical Support

> -----Original Message----- > From: Susan C. Underhill [SMTP:susan.underhill@UTORONTO.CA] > Sent: Tuesday, September 15, 1998 12:25 PM > To: SPSSX-L@UGA.CC.UGA.EDU > Subject: select cases (fwd) > > I am getting the same message as Bas here is but for a > totally different procedure. I am using SPSS 8.0 also and I > am trying to run a Kruskal-Wallis H test. > > Forwarded Message: > From: Bas van den Heuvel > <b.h.vandenheuvel@NOSPAM.STUDENT.UTWENTE.NL> > Date: Sun, 13 Sep 1998 07:26:59 -0400 > Subject: select cases > To: SPSSX-L@UGA.CC.UGA.EDU > > I am using SPSS 8.0 > When i do > - select cases, variable > 0, option unselected cases are > filtered > - weight cases by the same variable as i use for filtering > - and then i create a basic table i get following warning > > >Warning # 3211 > >On at least one case, the value of the weight variable was > zero, > negative, > >or missing. Such cases are invisible to statistical > procedures and > graphs > >which need positively weighted cases, but remain on the > file and are > >processed by non-statistical facilities such as LIST and > SAVE. > > but when i do select cases option unselected cases are > deleted with the > same select condition and the same weight condition. I > didn't get the > warning. > > All results in both situiation are the same. > > What am i doing wrong. > > Bas > > > > ************************************************************ > Susan C. Underhill, Senior Research Officer > Institute for Human Development, Life Course & Aging > 222 College St., Ste.106 > Toronto, ON M5T 3J1 > PH: (416) 978-5968 FAX: (416) 978-4771 > ************************************************************