I am doing a binary logistic regression , the outcome is dichotomous the answer is yes or no, I have several dependent variables some of them are dichotomous and others are categorical.
When I run the univariate binary logistic regression on spss taking the outcome versus the covariate without making categorization of the covariate the result was significant.
However, when I run it another time but with categorization of the covariate the result was insignificant. I have changed the reference from the last to the first but also the result was insignificant.
Would you please help me with this problem why it occured and how to solve
I don't understand what you are trying to do but the reason appears to me to be that in one case you ran the covariate as it came. In the other you categorized it. It is well-known that categorizing a variable throws away information. Hence those two covariates are not the same. Changing the reference variable is equivalent to changing a baseline. The other regression coefficients are now compared to a different base but (a-b) = :--(b-a) so the probability statements don't change. For a ref see: Applied Linear Statistical Models | Michael H. Kutner, William Li, Christopher J. Nachtsheim, John Neter | download (b-ok.cc)
Best, David Booth
I don't understand what you are trying to do but the reason appears to me to be that in one case you ran the covariate as it came. In the other you categorized it. It is well-known that categorizing a variable throws away information. Hence those two covariates are not the same. Changing the reference variable is equivalent to changing a baseline. The other regression coefficients are now compared to a different base but (a-b) = :--(b-a) so the probability statements don't change. For a ref see: Applied Linear Statistical Models | Michael H. Kutner, William Li, Christopher J. Nachtsheim, John Neter | download (b-ok.cc)
Best, David Booth
Are you representing your categorical variable(s) as a set of dummy variables?
Hello Islam,
I, too, have a problem fully comprehending your query.
If your independent variable was continuous, and you subsequently categorized it (and either you or the software appropriately recast it as k - 1 dummy variates, as suggested by David L Morgan ), then the discrepancy is attributable to the loss of information, as suggested by David Eugene Booth . If the categorization was incorrectly done, then the discrepancy is attributable to incorrect recasting of that IV.
However, if your independent variable was genuinely nominal (e.g., 1 = truck, 2 = airplane, 3 = roller skates, 4 = bicycle) and you just treated the numbers as if they were a continuous scale, then any discrepancy between treating such a variable as a continuous variable vs. a categorical variable is meaningless: the continuous version of the scores has no meaning and shouldn't have been evaluated.
Good luck with your work.
- Badges
- Science topic
- Similar topics
- Biological Science
- Kinesiology
- Running