Can you please explain the meanings of columns, especially “logDet” one? I haven’t found them neither in manual, nor in Holmes et al., 2012
Can you please also explain what does it mean “-nan”? I see that my Laplace-values haven’t gone to the minimum, oscillating instead. Does “-nan” mean that analysis was closed violently, because 7 is the maximum possible number of classes?
My apologies for asking, this calculation took literally a week on our computer, so I’m trying to understand, what went wrong, before to do a new try)
The logDet column is the log of the determinant. It’s been a while since I looked, but I think it was outputted in Chris Quince’s original code. What you really care about is K and Laplace and you want Laplace to hit a minimum. get.communitytype will continue on for several iterations after it finds a new minimum.
How many samples do you have? Are these OTUs, phylotypes, or ASVs?
My attention was sticked to that logDet column, cause it first positive and then became negative. I thought maybe there is some meaning in that. But maybe not)
I’m not sure I can securely trust my Laplace minimum, cause instead of finding one minimum, it oscillated and found the second, smaller, minimum, than after 2 more divisions stopped. What if it should oscillate more and reach best minimum e.g. when K=9? Is 7 the maximum possible number of classes?
Do you know what does an abbreviation “-nan” mean?
There are 130 soil samples from 4 horizons of the same soil, some of them are large 250 mg samples, some of them small 10-50 mg. There are OTU’s 97% (~19 000 OTU’s in all dataset).
I thought next time to try to analise large and small samples separately, or to analise only mineral horizons without humus horizon…
I’m not sure why you’re getting -nan - there’s no limit on the value of K. My only thought is that your samples are so diverse that it’s hard for the algorithm to find good types. I wouldn’t suggest getting fewer samples, rather you might want more. I think you should probably also try with genus-level phylotypes. I’m not sure what you mean by large and small samples - you want them all to have the same number of sequences (use sub.sample if you aren’t already).