I am working on some research work of women empowerment in Pakistan using demographic data, I have to re-code all explanatory variables categorically. Which are the compulsory thing which I should check before applying the logit model, and which should be applied after running it?
Make sure the dependent variable is binary. I guess the dependent variable is Women Empowerment. You can use the WORLD BANK convention for the calculation of empowerment. Based on the index, you can create a threshold below which you may say there is no empowerment (Assign 0) and above that threshold you may say there is empowerment (Assign 1). Kindly do a correlation test of all the variables and reject a variable if the coefficient of correlation is greater than 5%. E.g. if the correlation between age and experience is greater than 0.05 you can reject one of the variables (either AGE or experience).
Thanks brother,
my research is really related to omen empowerment.
Regards
atta
Dear Atta Muhammad Asif,
Edward is correct in stating that the dependant variable needs to be binary for a logistic regression. However, if you have a variable measured at the interval/ratio level, I see no need in recoding it to a binary variable: with that procedure you would loose a lot of information. Also, I see no problems with adding multiple explanatory variables that are correlated. Perfect correlations - or close to it - would impose a problem of (multi)collinearity, but there is no problem otherwise.
Finally, there is no need to recode the explanatory variables in logistic regression to categories, as you seem to imply in your original question.
Best,
Rense
- Badges
- Science topic
- Similar topics
- Philosophy
- Esthetics
- Interpretation