ANCOM RESULTS (W values)

[Jibda (Jibda del Pilar Carreño)]

Hi,

I want to know if a W value of 1 in the ANCOM results is significant. I don´t know what is the “minimum” value of W that I can take as significant.

Thanks

[ebolyen (Evan Bolyen)]

Hey @Jibda,

See this post by @mortonjt for more info about W.

In short, a 1 is not a good result, only 0 could be worse. W is a count of tests that “succeed”, so it’s not a proportion like you might hope.

Sorry, but hope that helps!

[Jibda (Jibda del Pilar Carreño)]

Thanks: slight_smile :. I always appreciate your quick response. I have read about other issues about 0 values but I have troubles understanding what minimal W value can I take as meaningful. Can you help me with this?

[ebolyen (Evan Bolyen)]

Hey @Jibda,

Since the maximum W value depends on the number of features being compared, the “minimum” W value doesn’t quite exist either.

What you are mostly looking for is a natural break in the W values which is what the volcano plot tries to show. For example, in this plot there’s two obvious points which separate far from the rest. But there’s still nothing like a “p-value” to demonstrate that those really are significant, they are just most dissimilar from the other W values.