dear allstat,
i have a logistic regression model for prediction purposes with predictors
a, b, c, d etc. suppose my study hypothesis states that predictor -a- is not
necessary, because the model based on the rest of the predictors classifies
just as well.
question is, how do i prove this statistically?
i am faced with the problem that even if the inclusion of -a- does not
significantly improve the model, it is expressed in terms of a high p-value
in a wald or a likelihood ratio test, and i do not want to interpret a high
p-value as proof in favor of the null hypothesis.
any hints to a methodologically acceptable solution pathway or to sources
where this problem is dealt with will be highly appreciated.
laszlo
|