I read different things in every book or website. My model concerns a very simple model with a relatively small sample (n=130).
2 predictors (both coded 0-1)
1 interaction between these predictors
1 dependent variable (coded 0-1)
KR, Thijs
Since your predictors are binary, the normality assumption is not needed.
Another issue you need to consider is multicollinearity between them. You'll need to use Spearman correlation coefficient.
I just found a nice primer on logistic regression by Stoltzfus (2011). Here's what it says about the basic assumptions:
- Basic assumptions that must be met for logistic regression include independence of errors, linearity in the logit for continuous variables, absence of multicollinearity, and lack of strongly influential outliers. Additionally, there should be an adequate number of events per independent variable to avoid an overfit model, with commonly recommended minimum "rules of thumb" ranging from 10 to 20 events per covariate. (p. 1099, in the abstract)
See more extended discussion of each of these on p. 1101. Regarding the rule of thumb for events-per-variable, see Mike Babyak's nice article (2nd link).
HTH.
http://onlinelibrary.wiley.com/doi/10.1111/j.1553-2712.2011.01185.x/pdf
https://people.duke.edu/~mababyak/papers/babyakregression.pdf
Take a look at Jason Osborne's "Best Practices in Logistic Regression."
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Coding