topic Re: multivariate analysis with categorical variables in Statistical Procedures

multivariate analysis with categorical variables

hein68 — Fri, 16 Jul 2021 14:39:18 GMT

What statistical test would you use for this analysis?

Dependent variable: response rate (Y/N)

Independent variable: U.S. state (there is a list of 10 states)

Control for recruitment method: categorical variable with 2 categories

Thanks!

Re: multivariate analysis with categorical variables

PaigeMiller — Fri, 16 Jul 2021 14:43:24 GMT

There are probably a number of ways of handling this in SAS, but PROC LOGISTIC is where I would start.

Re: multivariate analysis with categorical variables

Reeza — Fri, 16 Jul 2021 15:10:22 GMT

General overview of Logistic regression

https://www.listendata.com/2013/04/logistic-regression-analysis-with-sas.html

SAS Specific example - I suggest you work through this before trying it with your data

https://documentation.sas.com/doc/en/pgmsascdc/9.4_3.4/statug/statug_logistic_examples02.htm

@hein68 wrote:

What statistical test would you use for this analysis?

Dependent variable: response rate (Y/N)

Independent variable: U.S. state (there is a list of 10 states)

Control for recruitment method: categorical variable with 2 categories

Thanks!

Re: multivariate analysis with categorical variables

Ksharp — Sat, 17 Jul 2021 11:03:16 GMT

PROC CATMOD is designed for category variables anaylsis .

Re: multivariate analysis with categorical variables

StatDave — Fri, 30 Jul 2021 15:28:08 GMT

You haven't said exactly what you want to test or estimate, which is a vital thing to state when asking for a recommendation on an analysis method. Assuming that what you want to do are pairwise comparisons among the states, then this could be most easily done using PROC LOGISTIC and the LSMEANS statement. For example, these statements would provide multiple comparisons among the states adjusted for recruitment. This assumes that the effect of recruitment is the same across the states... if not than the interaction needs to be included. The ILINK option gives the estimated adjusted probability of RESPONSE="Y" for each state. The DIFF option provides a table of tests comparing the states.

proc logistic;
class state recruitment / param=glm;
model response(event="Y")=state recruitment;
lsmeans state / ilink diff;
run;