I have six non-linear independent continuous variables that I would like to include in my logistic regression to estimated adjusted ORs.
I performed the logistic regression using this syntax code:
__
LOGISTIC REGRESSION VARIABLES VariableX
/METHOD=ENTER Factor1 Factor2 Factor3 Factor4 Factor5 Factor6
/PRINT=GOODFIT CI(95)
/CRITERIA=PIN(0.05) POUT(0.10) ITERATE(20) CUT(0.5).
____
But the output results look messed up, I get numbers like "7937E.+12" and zeros for Exp(B) for the different indepedent factors.
Any solution? What is the right way to perform multivariable regression analysis for such dataset?
Hi Salma,
Try to use the variables transformation as in the Link:
https://medium.com/analytics-vidhya/advance-nonlinear-variable-transformations-ecedf3f8709e
hi Salma.
Chapter 4 of categorical data analysis book written by Alen Agresti (third edition) is about Logestic Regression. You can study more about that and also it has some codes with R software maybe helps you . Good luck
To follow up with @Salma excellent suggestions I would suggest that you use the same rhs that you may use with any linear IVs. It doesn't matter because the IVs in the standard linear model are assumed fixed. The wild funny numbers are written in scientific notation because programming professionals know regression coefficients can vary between large positive and large negative numbers and this allowed the programer to use one format for all size regression coefficients. Google scientific notation for full details. With best wishes, David Booth
Apologies to @Samira for confusing the names. Best wishes, David Booth
David Eugene Booth Thank you so much for Dr. Booth's always great answers
- Badges
- Science topic
- Similar topics
- Condensed Matter Physics
- Zero