1
$\begingroup$

I am running a logistic regression analysis with binary variables on SPSS:

dependent variable:

  • preterm birth (Y/N)

independent variables:

  • hypertension (Y/N), diabetes (Y/N), C section (Y/N), younger than 20 (Y/N), older than 35 (Y/N)

3 questions:

  1. Is it necessary to test the correlation between the independent variables first prior to running the logistic regression model?

  2. If so, how do I test it? Pearson's chi-square?

  3. What are the implications on the logistic regression model?

$\endgroup$
  • $\begingroup$ You have a bigger problem. The dependent variable and the independent variables are improper. They assume discontinuous biology and suffer phenomenal information loss. $\endgroup$ – Frank Harrell Feb 18 '15 at 12:41
2
$\begingroup$

This should be a comment but as I can't comment yet so I leave as an answer....

"younger than 20 (Y/N), older than 35 (Y/N)"

If you have the actual age, I think using age is better, 20 and 35 sounds like a random cutoff to me.

$\endgroup$
1
$\begingroup$

Why do you think you need to test correlation of independent variables (covariables) prior to logistic regression? No, there is no such need. If the covariables are highly correlated, there might be problems with colinearity, just as in linear regression. But there is no need to "test" for this prior to modelling.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.