How do I tell if there are problems with the importance sampling?

Look at the output on the screen, or use out=amread(dbuf,"resamp");, to check how many resamples were required to accept $m$ draws from the asymptotic distribution. The number must be between 1 and $m$. If the number is $m$ (or only 1 or 2 less than $m$) then you may want to consider resetting the globals to improve the importance sampling. It may suffice to increase _AMsn so that the number of draws from the asymptotic distribution is increased. Alternatively, you can try changing the asymptotic distribution to the student $t$ distribution. You can increase (but probably do not decrease) _AMsfac. It may also be helpful to use or reset the priors. In our experience, the only problems with resampling come with data that have largely nominal or heavily skewed variables or other features that severely violate the model assumptions.