Trouble with Ordinal Regression / Correlations

Roxxanne — Tue, 06 Jul 2021 18:19:29 GMT

Hello,

I am struggling to determine the correct code and/or procedure for my dataset. My dependent variable (i.e. Equity_1L) is discrete and not normally distributed. I have multiple categorical independent variables (i.e. age, sex, education, income, relationship, cancer and healthcare - only cancer and healthcare are dichotomous, the others have multiple levels). My main analysis was to determine whether the median value of my dependent variable differed from a known value. I completed this and wanted to check for correlations, after doing some research, I determined that an ordinal regression analysis would be the best fit for my data. I attempted this code and received an error:

proc logistic data=WORK desc;

model Equity_1L = Age Sex Education Income Relationship Cancer Healthcare;
run;

The error stated my variables need to be either numeric or specified by a class statement, so I moved all non-dichotomous to the class statement:

proc logistic data=WORK.THESIS desc;
model Equity_1L = Cancer Healthcare;
class Age Sex Education Income Relationship;
run;

I still receive an error saying the same thing. Basically, I am unsure if I am suppose to re-code these variables as dummy variables or try a different analysis. I did perform a 1 on 1 Wilcoxon analysis with each independent variable with my dependent variable to check for correlations however, I wasn't sure if this was the proper way to do it, as I thought all correlations should be checked for at once (just in case there are interactions). Thank you for your help!

Re: Trouble with Ordinal Regression / Correlations

Rick_SAS — Tue, 06 Jul 2021 19:12:10 GMT

Put the CLASS statement before the MODEL statement.

topic Trouble with Ordinal Regression / Correlations in Statistical Procedures

Trouble with Ordinal Regression / Correlations

Re: Trouble with Ordinal Regression / Correlations