comparing individual and groups in datasets in the same
in [SAS]
|
Prev: Intentionally specifying an unclosed %DO loop in a macro XXXX
Next: C or Java code in Extension Nodes
From: Arthur Tabachneck on 13 Feb 2010 23:43 John, If you are asking the statistical question, I'd think the assumption of INDEPENDENCE would be violated by definition. If your question isn't statistical, and you're only interested in seeing the comparisons, then simply code the desired group into a new group and output both the original and recoded records. Art ------- On Feb 13, 7:55 pm, johnthomas75 <johnthoma...(a)gmail.com> wrote: > If I want to compare 4 groups where 1 of the groups is the composite > of 2 of the other groups and the 3 other groups are individual groups > then how do I code that into an analysis so that I compare the > multiple groups to the single groups. > > i.e. groups= 2, 3, 4 and 3&4 > I want to compare 2 to 3 to 4 to 3&4 in proc freq? > > how do I code it?
From: Arthur Tabachneck on 13 Feb 2010 23:39 On Feb 13, 7:55 pm, johnthomas75 <johnthoma...(a)gmail.com> wrote: > If I want to compare 4 groups where 1 of the groups is the composite > of 2 of the other groups and the 3 other groups are individual groups > then how do I code that into an analysis so that I compare the > multiple groups to the single groups. > > i.e. groups= 2, 3, 4 and 3&4 > I want to compare 2 to 3 to 4 to 3&4 in proc freq? > > how do I code it?
From: Jim Groeneveld on 16 Feb 2010 04:53
Hi John / Thomas, I suppose your groups are indicated as distinct variables, let's say they are called Group2, Group3, Group4 and Group34. If Group34 is a direct combination of two other variables (from the three), in this case Group3 and Group4, but you initially would not know which ones, you may study their relationship by producing cross tables indeed. You would produce a cross table of Group2, Group3 and Group34, one of Group2, Group4 and Group34 and one of Group3, Group4 and Group34. From the results you may tell which independent variables determined the composite variable (if the relationship is not complex mathematical with many empty (zero freq) cells). Why wouldn't you know from which variables the composite has been deduced? PROC FREQ DATA=.............; TABLES Group2 * Group3 * Group34 / Group2 * Group4 * Group34 / Group3 * Group4 * Group34 ; RUN; Regards - Jim. -- Jim Groeneveld, Netherlands Statistician, SAS consultant http://jim.groeneveld.eu.tf On Sat, 13 Feb 2010 16:55:01 -0800, johnthomas75 <johnthomas75(a)GMAIL.COM> wrote: >If I want to compare 4 groups where 1 of the groups is the composite >of 2 of the other groups and the 3 other groups are individual groups >then how do I code that into an analysis so that I compare the >multiple groups to the single groups. > >i.e. groups= 2, 3, 4 and 3&4 >I want to compare 2 to 3 to 4 to 3&4 in proc freq? > >how do I code it? |