Hey - it’s great to see this command added. I’ve used the DMM software a number of times in the past, and it was always a pain to convert from a shared file to what their software reads. I had a few questions about the commands, since the wiki at the moment reads like “The minpartitions parameter is used to …”.

I’m assuming that the min/maxpartitions parameters refer to the number of Dirichlet components to test, and that mothur repeats from minimum to maximum. Is this correct?

Also, I’m not sure what the optimizegap option does (according to the wiki it does ‘…’). In the original software you could change the seed value used for calculations and this could (slightly) influence the fit of the model. Is this what the optimizegap option refers to? If not, exactly how does it optimize the results?


You mean you want documentation!? :slight_smile:

So minpartitions is the minimum number of metacommunities to test and maxpartitions is the maximum number to test. By default min is 5 and max is 100. So it will test at least 1-5 partitions. The optimizegap tells the command when to stop (you don’t really want 100 partitions). By default the value is 3. So if you do 1, 2, 3 partitions and 3 has the lowest Laplace value, it will do at least 3 more partitions before stopping. If it finds a new minimum Laplace value in those 3, the count will restart. Our experience has been that the decrease/increase in Laplace values is not monotonic. Also, we do occasionally see some run-to-run variation in the number of metacommunities generated as there is some randomization involved.

We’ll work on better documentation :slight_smile: