In the procedure of backward selection, after dropping the variable with the largest p-value, should I do likelihood ratio test of the reduced model with the original model first, and then check the variable with the largest p-value again or should I directly look for that variable without the likelihood ratio test?
Asked
Active
Viewed 20 times
1
-
This is another question that might get a better answer at stats.stackexchange.com. $\qquad$ – Michael Hardy Feb 19 '16 at 19:44