Nmds output -nan

1

NMDS no results for every large number of samples, e.g., 60K samples.

all are -nan in the output files, no error message

Thanks,

Jianshu

2

I suspect you have too much data to run nmds or you have really large distances that it can’t work with.

I’m curious why you would want an ordination with 60k samples. Isn’t it just going to look like a big blob of stuff? It’s hard for me to imagine why this would be useful.

Pat

3

Hello Pat,

I am visualizing genomic distance, each sample is now a genome, distance is genomic distance approximated by MinHash. Distances are like, 0.2 to 0.85, highly uneven distributed, towards to large end.

Thanks,
Jianshu

4

Suggestions for any other way to visualize it is welcomed (linear dimension reduction only).

Thanks,

Jianshu

5

I always encourage people to first ask what the question is that they’re trying to answer then go from there. If your inclination is an ordination, then you’re probably trying to show how similar/dissimilar things are to each other. Can you bin genomes by some other variable and represent the distance within or between each category as a box plot? Were the genomes sampled over time - you could put distance to day 0 (or day -1 or whatever) on the y-axis and time on the x-axis.

Ordinations really lose their value after about 20 samples IMHO.

Pat