Hi.

Divide the data in bins that are roughly equally sized and calculate the % of "hits" of your dependent measure within each bin. Correlate with the inverse logodds of the mean of the predicted logits within each bin. I think I have a function that does this somewhere.

Thank you. I'm going to try this.

If you have a function to do this, I would be happy if you can share it with me.

Here's a function that calculates probabilities in bins whose size either are determined on the basis of quantiles of the distribution of predicted values or is set. It also returns mean predicted values and mean residuals per bin. It can also generate per-bin confidence intervals on the basis of simulation and provide per-bin bootstrapped proportions and CIs, if requested. You can provide a lrm, glm or glmer model. It was a while since i wrote it so I am not 100% familiar with it but it should work fine.

Oh, it can also calculate the above mentioned binning statistics on the basis of some predefined type of binning such as fixed or random effects in the model (e.g., gender, subject or item differences). Pretty neat. It seems that it requires other functions not availible in any package that I have written myself, though.

Thanks a lot! I'm working on it.