turn on suggestions

Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.

Showing results for

Find a Community

- Home
- /
- Analytics
- /
- Stat Procs
- /
- dmreg / logistic regression interpretation

Topic Options

- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

05-21-2011 12:39 PM

Hi all,

the dmreg procedure returns under "Summary of Stepwise Selection" a "Score Chi-Square". Given two effects a and b which are binary i.e. the value of a and be is always 0 or 1. The effect a has a significantly higher score Chi-Square than b. How can it be, that the parameter estimates for b (shown under "Analysis of Maximum Likelihood Estimates") is larger than the one of a?

The reason i would like to understand this is, that i am doing a scoring of customers. In addition to the score value per customer i would like to also show the main driver for the scoring per customer, i.e. the effect which has the strongest impact on driving the score high. The first idea was to identify the main driver by taking the largest contribution to the score (product of parameter estimate multiplied with effect value). However, it turned out, that effects, which i expected to be often shown as the main driver due to a high Chi-Square score, get overruled by other effects with smaller Chi-Square score. How can i know the order of the importance of the effects? I thought by the chi-square, but now i am not so shure any more...

Looking forward to your comments.

Georgios

the dmreg procedure returns under "Summary of Stepwise Selection" a "Score Chi-Square". Given two effects a and b which are binary i.e. the value of a and be is always 0 or 1. The effect a has a significantly higher score Chi-Square than b. How can it be, that the parameter estimates for b (shown under "Analysis of Maximum Likelihood Estimates") is larger than the one of a?

The reason i would like to understand this is, that i am doing a scoring of customers. In addition to the score value per customer i would like to also show the main driver for the scoring per customer, i.e. the effect which has the strongest impact on driving the score high. The first idea was to identify the main driver by taking the largest contribution to the score (product of parameter estimate multiplied with effect value). However, it turned out, that effects, which i expected to be often shown as the main driver due to a high Chi-Square score, get overruled by other effects with smaller Chi-Square score. How can i know the order of the importance of the effects? I thought by the chi-square, but now i am not so shure any more...

Looking forward to your comments.

Georgios