Stepwise Regression with nominal predictor variable

Question

Hi all,

I would like to carry out a stepwise regression with a continuous variable as y and mixed continuous and nominal predictors. To get an idea of the programme (new jmp user; JMP Pro 13), I downloaded the Hollywood Movies dataset from the Sample Data Library and followed all suggested steps of the “Fitting-Linear-Models”-pdf (version 12), pp. 246-248.

The intercept term and the first three terms (Rotten Tomatoes Score, Audience Score, Genre) of the estimates table I produced were identical with that of the pdf (p. 248), but the values of all other Genre terms differ (see attached).

Does anyone have an idea why that might be the case? (I would just like to make sure that everything is in order before tackling my own dataset.) Thanks for all help.

Best wishes,

Manni

Hollywood_stepwise_question.JPG

Mark_Bailey · Answer

I reproduce the results published in the book. Are you sure that you followed the steps exactly?

If so and you still get different results, contact JMP Technical Support. at support@jmp.com and explain your issue.

Manni · Answer

Thanks for your answer. I ran it with log(World Gross), but this didn’t do the trick. The values that were correct before are off as well now.

Hollywood_stepwise_question2.JPG

Mark_Bailey · Answer

I followed the instructions in JMP 12 "Fitting Linear Models" starting on page 246 using JMP Pro 13. I see no difference.

Your picture does not agree at all because youi are not modeling Log( World Gross ) but just World Gross.