Why statistic test should I use?

Question

[Matheus]

I"m comparing two sets( a and b) of genomes in some metrics(genome size, gc content, number of gene). A is a set of about 130 genomes, and B(30 genomes) is a subset inside A that have some characteristics that I want. I have calculated the average of those metrics I both as sets and that I want to know if those averages are significantly different. Which test should I use? I am thinking about one sample t test, but my data is not normally distributed.

[ATpoint]

Without seeing the data, why not just give the Wilcox test a try? It makes no distributional assumptions and 130 vs 30 is big enough of an n to get significances with it. It tests ranks, not means though.

[dariober]

B(30 genomes) is a subset inside A

Regardless of the test you choose, shouldn't you compare B vs (set of A genomes without B genomes)? I can't really tell why, but comparing a subset vs the whole seems odd to me.

Answer 1

[shelkmike]

I think you can do a Mann-Whitney test. Strictly speaking, it does not compare means (the null hypothesis is slightly different from the equality of means), but it probably is still suitable for you.