topic Ordinal predictors in Quantreg in Statistical Procedures

Ordinal predictors in Quantreg

Giampaolo — Mon, 18 Oct 2021 17:24:57 GMT

Dear SAS experts,

I have two variables income (4 levels) and education (3 levels) that I would like to use as predictors for quantile regression. I coded them as class variables but I believe it would be more correct to define them as ordinal. Is it possible to define independent variables as ordinal in Quantreg?

Thank you

proc quantreg data=Mydata CI=resampling order =internal;
class Income education mammography;

format Income Income. Education edu.;
model DependentV = Income Age|Age education mammography / diagnostics quantile=0.25 0.5 0.75 covb
plots= quantplot ;
test income/LR rankscore wald;
ods output parameterestimates = PMSQUANT;
run;quit;

Re: Ordinal predictors in Quantreg

Rick_SAS — Mon, 18 Oct 2021 17:48:50 GMT

No. Some SAS regression procedures enable you to specify an ordinal variable for the RESPONSE variable, but I don't think ordinal INDEPENDENT variables are supported on the CLASS statement.

Some people code ordinal explanatory variables (Likert variables) as evenly spaced integers, such as -1, 0, +1 for "disagree", "neutral", and "agree". You would then treat the variable as a continuous variable and NOT put it on the CLASS statement.

Re: Ordinal predictors in Quantreg

Giampaolo — Mon, 18 Oct 2021 18:28:04 GMT

Thank you!