Interpreting the significance output of libshuff

Hi Pat,

I have used mothur to test whether my samples have different community structures, using the libshuff command.

Here is (part of) the output from mothur.
Comparison dCXYScore Significance
S1.fasta-S2.fasta 0.00004858 1.0000
S2.fasta-S1.fasta 0.00132960 <0.0001
S2.fasta-S3.fasta 0.00007639 <0.0001
S3.fasta-S2.fasta 0.00004697 <0.0001
S2.fasta-S4.fasta 0.00005334 <0.0001
S4.fasta-S2.fasta 0.00001421 0.0003

Having read the information about libshuff on the website, am I correct to say that all the comparisons shown here (S1 vs S2, S2 vs S3, S2 vs S4) indicate a significant difference in the community structure of the samples? I just wanted to check because there are very contrasting significance values for when S1 was compared to S2, and when S2 was compared to S1. I also remember seeing a poster about S-libshuff detecting subsets when both p-values are significant. Can you clarify this please?

In addition, is the significance value equivalent to the p-value, and what is the best way to quote a p-value for each pair of sample, as there are two significance value for each pair. Thanks for your help.


Your interpretation is correct. I showed in my 200X ISMEJ paper that libshuff cannot detect subsets. The P-value of 1.000 is weird. I guess I’d report the minimum P-value for each set of comparisons.