LDALY@IRLEARN.BITNET (LESLIE DALY PhD) (01/31/90)
Lisa King <LMJY@CORNELLA.BITNET> had a query re TABULATE when it produced two
different sets of answers with the following two sets of code:
PROC TABULATE; PROC TABULATE;
CLASS V10A V28; AND CLASS V10A V10B V28;
TABLE V10A,V28; TABLE V10A V10B,V28;
The answers differed in the part of the table giving N of V10A by V28.
The reason is that V10B is a CLASS variable that has missing values for
2 observations where V10A is valid. In the first case the TABULATE is
correct and gave what was expected. In the second case the *observations*
with the missing values were omitted from ALL tables requested. 'A missing
value in a class variable causes the observation to be ignored' Basics manual
p 1098. This happens even if the particular class variable does not form
part of the crosstabulation (to borrow SPSS terminology) in question. I have
always found this to be a bit of a bother but I think there are good reasons
for it in terms of consistency of table results.
--------------------------------------------------------------------
Leslie Daly Ph.D. (Sex: male!) Telephone: +353-1-693244 ext 6315
Lecturer in Biostatistics, Network: LDALY@IRLEARN.BITNET
Department of Community Medicine
and Epidemiology,
University College Dublin,
Earlsfort Terrace,
Dublin 2
IRELAND.
--------------------------------------------------------------------