0

Hey I'm doing a logistic regression looking at predicting if someone will commit one type of crime or another based on 11 ordinal variables (each variable is 0-2). I'm having some trouble interpreting the results of my SPSS output and wouldn't mind some advice. I've given some images of the output to make this clearer. At Step 0, the constant is not significant, and also only variable 4 is significant. I know this is poor, but I'm still not quite sure what this means (e.g. maybe its inappropriate to run the analysis at all. In the final image, when the variables are added to the equation, variables 2, 4, 5 and 8 are significant, which means they might be good predictors? My stats knowledge is based on ideal world scenarios where everything is significant, so I am finding it difficult to understand this data, so any help is appreciated.

Step 0

enter image description here

enter image description here

KyleKatarn
  • 1
  • 1
    Check this page for how to interpret the logistic regression SPSS output https://stats.idre.ucla.edu/spss/output/logistic-regression/ – Heteroskedastic Jim Jul 11 '18 at 11:17
  • 1
    Looking at your last table it is clear you have separation so you need to follow-up with some of the questions and answers in the thread here on separation. Try this one https://stats.stackexchange.com/questions/11109/how-to-deal-with-perfect-separation-in-logistic-regression first. Without knowing more about your scientific question and your variables that is as far as we can go. – mdewey Jul 11 '18 at 13:40

1 Answers1

1

You don't care whether step 0 is significant. This is the equivalent in OLS regression of the intercept only model (i.e. the total sum of squares). It's not interesting.

It's not poor if predictors are not statistically significant, if that's the truth. You don't seek significance. You seek truth when you do statistical analysis.

I don't know what an ideal world scenario is where everything is significant.

Also, don't use the classification table to judge your model. It can vary between useful and deceptive.

Finally, this is a question about logistic regression interpretation, it's not a question about SPSS. It doesn't matter what program you use to estimate your regression. (If you ask a question about SPSS your question will be closed, this is a forum about statistics, not software.)

Jeremy Miles
  • 13,917
  • 6
  • 30
  • 64