<div>Hi,</div>
<div> </div>
<div>You may have solved this already, but just for the record:</div>
<div> </div>
<div>ADMB has no special provision for finite mixture models, except that</div>
<div>its fast and stable funciton minimizer clearly is an advantage </div>
<div>when fitting such models. The random effects module is only for</div>
<div>continous random effects. To implement a finite mixture model</div>
<div>you have to write the C++ code directly. A simple example</div>
<div>of a Gaussian mixture is:</div>
<div> </div>
<div>DATA_SECTION<br> init_int n // Number of observations<br> init_vector x(1,n) // data <br>PARAMETER_SECTION<br> init_number p1 // Mixture probabilities<br> init_number p2<br> init_number mu1 // Mean values<br>
init_number mu2<br> objective_function_value g // Log-likelihood<br>PROCEDURE_SECTION<br> for(int i=1;i<=n;i++)<br> g += log(p1*square(x[i]-mu1)+p2*square(x[i]-mu2));<br> g *= -1; // ADMB does miminization!<br>
</div>
<div>This code does not enforce the constraint p1+p2=1 or deal</div>
<div>with the identifiability issue regarding the mu's.</div>
<div> </div>
<div>Hans</div>
<div> </div>
<div>>I am new to ADMB and find the approach it adopts very interesting. One thing<br>>I would like to know is: does ADMB handle models in which the random effects<br>>are not parameterized as continuous distributions (normal, gamma, etc.) but<br>
>as discrete categories (e.g. finite mixture)? Handling finite mixture is<br>>computation intensive and tends to be very slow for most statistical<br>>packages; it can be a great strength (if ADMB handles it) of the unique<br>
>approach adopted by ADMB.<br>><br>>I would appreciate some examples, if any are available.<br>><br>>Many thanks.<br>><br>>Best,<br>>Shige<br></div>