empiricalhist to simulate latent thetas

[Elia Emmers]

Hello Phil! Quick question. I've been using the empiricalhist=TRUE option to look at how the latent theta looks like. If I wanted to get a better approximation to this distirbution, would increasing the number of quadrature points be the only way? And, is there a way to have access to the table that contains the data on the 'E' step from the EM algorithm to try and fit a distribution to it?

[Phil Chalmers]

On Mon, Jul 24, 2017 at 3:07 PM, Elia Emmers <eliae...@gmail.com> wrote:

Hello Phil! Quick question. I've been using the empiricalhist=TRUE option to look at how the latent theta looks like. If I wanted to get a better approximation to this distirbution, would increasing the number of quadrature points be the only way?

Yes and no. Each empirical quadrature is really another estimated parameter in the model, so there comes a point where creating a smooth distribution is infeasible. This is generally why empirical histogram models have difficulty converging (that, and they aren't always so stable).

 

And, is there a way to have access to the table that contains the data on the 'E' step from the EM algorithm to try and fit a distribution to it?

Sure, you should be able to extract the component from the x@Internals$Prior[[1L]] slot. HTH.

 

--
You received this message because you are subscribed to the Google Groups "mirt-package" group.
To unsubscribe from this group and stop receiving emails from it, send an email to mirt-package+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.