I have a SPSS data set with 70 respondents with disease and 70 control respondents without disease. I also have a list of 18 symptoms that they could have. Is there a way to ascertain which symptoms contribute more to the diagnostic? Which statistc test should I use?


1 Answer 1


Simple logistic regression can be used when the outcome variable is binary. You can code having and not having a disease as 1 and 0 and the 18 symptoms will be your predictor variables. If you're planning to use SPSS, perhaps you might find this useful

  • $\begingroup$ With 18 features and 140 datapoints that seems to be quite likely to result in some overfitting. Also, using this to decide what to include in the model based on some of the things people sometimes use (e.g. p <x) tends to be a bad idea. $\endgroup$
    – Björn
    Nov 14, 2016 at 7:06
  • $\begingroup$ And what entrance method should I use? $\endgroup$
    – Andre
    Nov 14, 2016 at 15:25
  • $\begingroup$ If there is overfitting you could use logistic regression with the lasso, see stats.stackexchange.com/questions/212613/… $\endgroup$ Apr 28, 2017 at 22:50

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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