To get any sort of diagnostic plots, we need to be able to simulate responses from the mixture density. We can do it by simulating responses from the experts and then resampling those, or by using something like importance or slice sampling which requires evaluating the experts' densities. Then we should be able to exploit the diagnostics provided by mgcViz quite easily.
To get any sort of diagnostic plots, we need to be able to simulate responses from the mixture density. We can do it by simulating responses from the experts and then resampling those, or by using something like importance or slice sampling which requires evaluating the experts' densities. Then we should be able to exploit the diagnostics provided by mgcViz quite easily.