<p>Dave and Anders,<br></p><p></p><p>There are clearly things ADMB can do with mixed models that no</p><p>other software package can do. The only other package that have</p><p>the flexibility to easily allow Dave's beta-response example to</p>
<p>be implemented is SAS NLMIXED. But NLMIXED is slow and</p><p>numerically unstable relative to ADMB.</p><p></p><p>Then there are other model types where there is no competition:</p><p>- crossed random effects (except for the limited set of models implemented in R)</p>
<p>- nonlinear statespace models.</p><p></p><p>Anders's suggestion about publishing a paper in "journal of statistical software" is good.</p><p></p><p>Hans</p><p></p>