search

zdk123 / SpiecEasi

Sparse InversE Covariance estimation for Ecological Association and Statistical Inference
GNU General Public License v3.0
191 stars 67 forks source link

Network with no edges #232

Open minzooey opened 1 year ago

minzooey commented 1 year ago

Hi, Kurtz

I performed "spiec.easi" with "MB" using 6 samples including about 200 species. I know the number of samples is insufficient for network inference. I have already performed this analysis using 5 or 6 samples.

But, I suddenly ran into the error as follows:

> se.mb.lths <- spiec.easi(lths.phy, method='mb', lambda.min.ratio=1e-3, nlambda=100, pulsar.params=list(rep.num=100, ncores=4))

Applying data transformations...
Selecting model with pulsar using stars...
Fitting final estimate with mb...
done
Warning messages:
1: 1 job had warning: "35 parallel function calls did not deliver results" 
2: In pulsar(data = X, fun = match.fun(estFun), fargs = args, rep.num = 100,  :
  Optimal lambda may be smaller than the supplied values

As a result, I got a network with 0 edges.

zdk123 commented 1 year ago

Try lowering lambda.min.ratio even further. https://github.com/zdk123/SpiecEasi#troubleshooting

But you just have too few samples to get a good covariance estimate.