+ Reply to Thread
Results 1 to 1 of 1

Thread: Lousy model or case of complete separation...?!

  1. #1
    Points: 75, Level: 1
    Level completed: 50%, Points required for next Level: 25

    Thanked 0 Times in 0 Posts

    Lousy model or case of complete separation...?!


    I'm struggling to get my model to fit my data through logistic regression in SPSS.

    I've got 3 categorical predictors:

    Behandl (yes/no)
    Medicin (yes/no)
    Annons (5 subgroups)

    and one dependent variable (cured - not cured).

    I've built the model hierarchically in order to see which predictor makes the best fit.

    As far as I can see (striving for parsimony...) only the model with the predictor "behandl" significantly improves the model (0.012)

    HOWEVER, the crosstabulation box does not show any improved results regarding Overall percentage, it looks exactly as in step Step 0.
    Though if I add one of the other predictors ("annons") that has a non significant improvement on the model, it suddenly rises the Overall percentage. But the "annons" predictor produces ridiculously large standard errors. I've attached the output since it is somewhat difficult to explain.

    Does anyone know why this is happening?

    I've struggled with this for a week now. Would be ever so grateful if someone can help me out!
    Attached Files

+ Reply to Thread


Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts

Advertise on Talk Stats