MixedLM profile likelihood issue #2102

Closed
jseabold opened this Issue Nov 20, 2014 · 1 comment

Projects

None yet

2 participants

@jseabold
Member

The default example in the MixedLM notebook example does not work for the profile likelihood. It will be merged in #2097 mainly because the steps are instructive I guess. I don't know if it's just a hard problem, there's a bug, or if there's anything that can be done in the algorithm to restrict the covariance to be positive-(semi?)definite or to fix the likelihood calculation to account for this.

I also don't know why it works in this notebook other than the usual version/system differences.

@kshedden kshedden added a commit to kshedden/statsmodels that referenced this issue Nov 30, 2014
@kshedden kshedden Fix bug in profile_re, closes #2102 d8e22b9
@kshedden kshedden referenced this issue Nov 30, 2014
Merged

Mixed profile #2111

@kshedden
Contributor

Thanks for catching this. I think it is fixed now.

@jseabold jseabold closed this in #2111 Dec 2, 2014
@jseabold jseabold added a commit that referenced this issue Dec 2, 2014
@jseabold jseabold Backport PR #2111: Mixed profile
This should close #2102.  I did not correctly update the profile likelihood code after creating a class for managing the model parameters.

I also updated the notebook at:

http://nbviewer.ipython.org/urls/umich.box.com/shared/static/6tfc1e0q6jincsv5pgfa.ipynb
13eaf0c
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment