I am getting odd results from the bayesglm function in the ARM package. Running R 3.0.2 and downloaded the most resent ARM package. Every dataset I try runs, but the output is incorrect - I get a null deviance that is less than the residual deviance. I have tested this with datasets I have used for a class many times before and know what the results should be, but get the same result - coefficients are different and the null deviance is less than the residual deviance. I am not a Bayesian person - I just like to run this to compare to the standard GLM in logistic regressions. Standard GLM works fine and gives same/similar results as I have gotten in the past with the test data, but the bayesglm does not. Any suggestions? Peter Harrell Faculty Instructor NSOE Duke University A224 LSRC Building 919.613.827 [[alternative HTML version deleted]]