Calculate your model in SPSS. Does your model fit the data?

Question

Answer 1

The first box here gives us the ‘omnibus test of model coefficients’. This gives us an indication of whether or not the model with our independent variables fits the data better
(i.e. gives us a better prediction of individual scores) than the baseline model. We can find significance in the final column on this table, and we can see (significance is less
than .05) that the model is significant, which means that our model with the three predictors fits better than a model with no predictors.
The next box provides us with the Pseudo R square statistics. There are two measures, Cox & Snell and Nagelkerke. Both use a somewhat different formula, but both are equally valid. In this case Cox & Snell is .15, and Nagelkerke is .21. These numbers indicate modest improvement in fit over the baseline model (0-.1 would indicate poor improvement in fir, .1-.3 modest improvement, .3-.5 moderate improvement and more
than .5 strong improvement).
The next box is called the classification table, and gives us the comparison between predicted scores and the actual scores. We can see, for example, that 274 pupils who were predicted to fail by our model (with the three predictors) did indeed fail, while 138 were predicted to fail and in fact passed. In total, 68.6% of our predictions were accurate, which though far from perfect is a clear improvement over the baseline model,
where 53.2% of predictions were accurate.