I want to ask if the higher value of B is a matter of concern for the health of the hierarchical logistic regression model or not? I have a categorical variable and for two of the categories, I have got a B value +20 and another one is -22 for which the respective exp(B) are approx. 1070563626 and the other is 0.000.
Asked
Active
Viewed 195 times
2
-
2Search this site for "logistic regression separation" – kjetil b halvorsen Dec 06 '17 at 15:31
1 Answers
4
It might be and it might not. It might be a result of quasi-complete separation (which could be problematic) or it might just indicate that your independent variable is very strongly related to your dependent variable.
Peter Flom
- 94,055
- 35
- 143
- 276
-
Many thanks for the reply. Actually I missed it somehow that this variable is not significant in terms of p value. So what should I do as this may be a case of quasi-complete separation as you mentioned earlier? One way could be leave this variable out or actually categorizations may need reconsideration. – Muhammad Shahzad Hanif Dec 07 '17 at 05:40
-
If a B like that is not significant, it is certainly problematic. QC separation is certainly one possibility. Or a very small sample. – Peter Flom Dec 07 '17 at 12:45
-
Dear Mr. Peter, Would you please check and comment on another of the problems facing. Here is the link to that querry. https://stats.stackexchange.com/questions/318235/big-change-in-p-value-and-b-value-due-to-moderator-variable – Muhammad Shahzad Hanif Dec 12 '17 at 09:52
-
Dear Mr. Peter, A very warm x-mas greetings to you and family. Would you please care to advise on a problem being faced. thanks. https://stats.stackexchange.com/questions/320624/should-i-use-weighted-average-to-deal-with-category-size-problems-in-a-logistic – Muhammad Shahzad Hanif Dec 28 '17 at 09:11