How to fix Multinomial error in the output?

hafidwr — Fri, 29 Sep 2017 06:10:17 GMT

I have a problem. I made a multinomial logistic regression analysis using SAS (I'm new to this). A have a code like this:

proc print data=lasso;
run;

proc logistic data=lasso;
class Y (ref="3") X2(ref="1") X3(ref="3") X4(ref="4") X5(ref="3") X6(ref="3") X7(ref ="3") X8(ref ="3") X9(ref="3") X10(ref="2") X11(ref ="4") X12(ref ="3")/param=ref;
model Y = X1 / link =genlogit;
run;

In the output. My categorical predictor won't come out. Only X1 which is non categorical, that came out. What's wrong with this?

Re: How to fix Multinomial error in the output?

Rick_SAS — Fri, 29 Sep 2017 09:40:15 GMT

The classification variables X2-X12 do not appear in the model because you have not listed them on the MODEL statement. It sounds like you want:

MODEL Y = X1-X12 / link=genlogit;

As to the Y variable, each nonreference value of that variable get a separate set of parameter estimates. In the parameter estimates table you will see an intercept for Y=1, Y=2, etc. Similarly, a coefficient for X1 for Y=1, Y=2, etc.

Look at the output for this example and notice the second column displays the levels of Y to which the estimates are associated:

data Have;
call streaminit(123);
do i = 1 to 100;
   x1 = rand("Normal");
	y = rand("table", 0.3, 0.4, 0.2, 0.1);
	output;
end;
run;

proc logistic data=Have;
class Y (ref="3") /param=ref;
model Y = X1 / link =genlogit;
run;

topic Re: How to fix Multinomial error in the output? in Statistical Procedures

How to fix Multinomial error in the output?

Re: How to fix Multinomial error in the output?