Dear Mary Joy,

I would like to point out three main things:

1) The significance of the model is not necessarily important per se. In general, we are interested in the significance of each of the effects in the model. For example, the effect "rescour".

2) In addition, since this is a categorical (factor) effect in the model, one of the levels (in this case, rescour = 1) is set to an effect of "0". Thus , all odds ratio are compared to that level (again, rescour = 1). If the effect of "rescour" is significant, we would expect that at least one of the levels is different from another.

3) I don't know which software you are using, but, 2 things are unclear to me:

3.1) I would expect to see a single p-value for each of the 2 effects ("ascorecat" and "rescour"). These would have 1 and 5 degrees-of-freedom, respectively.

3.2) It seems that the table with odds ratios is shifted. For example, "_Iascorecat_1" is being used kind of like the header of that column, but this should be associated with an odds ratio (and all other statistics as the other levels have).

Thus, could you please provide more output from the software? Thanks , Nick

If these are actual dummy variables they are orthogonal so not colinear. This can occur for lots of reasons, including these are testing different things and it looks like the significance is estimated in different ways (Wald v Likelihood ratio), but it would be helpful to show the 2x6 contingency table (your Pearson Chi^2 should be close to the LR Chi^2). What may be happening is that your reference category has few cases so comparing it with any of the other categories is not powerful. Main thing: more information is needed to answer this question.

Thank you for your answers. :)