visualizing categorical-continuous interaction with proc plm

withpeople86 — Wed, 22 May 2019 02:00:49 GMT

I tried to visualize the predicted probabilities from the results of proc glimmix procedure using following codes.

proc glimmix data method=laplace noclprint;

class sex unemploy district;

model death (event=last)= age sex unemploy index unemploy*index/cl dist=binary solution;

random intercept /subject=district s cl type=vc;

covtest/wald;

store results; run;

proc plm restore=results;

effectplot slicefit (x=index sliceby=unemploy)/clm;run;

The results showed the graph, but the problem is that since the coefficient of interaction is very tiny, so the graph doesn't show the difference. What codes should I add to set the ranges of the axis X and Y. For example, the default range was 0, 0.5, 1 and 1.5 for x, but I want to change these figures to 0, 0.05, 0.1, 0.15, 0.2.

Thanks in advance.

Re: visualizing categorical-continuous interaction with proc plm

StatDave — Wed, 22 May 2019 18:39:14 GMT

You can specify the range on the Y axis using the YRANGE=(min,max) option in the EFFECTPLOT statement after a slash. While there isn't an equivalent XRANGE= option, you can specify the EXTEND=value option to extend the lines and X axis by value*(range/2), where range is the range of the X axis. See the descriptions of these and all the options in the EFFECTPLOT statement documentation.

topic Re: visualizing categorical-continuous interaction with proc plm in Statistical Procedures

visualizing categorical-continuous interaction with proc plm

Re: visualizing categorical-continuous interaction with proc plm