Rand_formula and group

[aston9090]

Hello,

Thank you for developing and maintaining the ANCOMBC package!

I have been trying to analyse microbiome data. In the data, there are 5 groups, and each group contains 2 species, meaning that there are 10 different species in total. I would like to compare the abundance of microbes across groups while accounting for species (which explains more than 50% of variations). I tried to run ancombc2 and got an error message. Please see below.

ANCOMBC2<-ancombc2(data=data, fix_formula = "Group", rand_formula = "1|Species", group= "Group" ,global = T, tax_level = "phylum")

Error in dimnames(x) <- dn :
length of 'dimnames' [2] not equal to array extent
In addition: There were 50 or more warnings (use warnings() to see the first 50)
Warning messages:
1: Model failed to converge with 8 negative eigenvalues: -9.1e-05 -1.0e-04 -1.0e-04 -1.0e-04 -1.0e-04 -3.4e-04 -3.5e-04 -3.5e-04
2: Model failed to converge with 2 negative eigenvalues: -8.1e-05 -8.1e-05
...

Any help and suggestions are greatly appreciated. Thank you!

[FrederickHuangLin]

Hi @aston9090,

I have reviewed the syntax you provided and have noticed the issue that may be causing problems.

It is important to specify the parenthesis for random_formula. For example, instead of rand_formula = "1|Species", it should be written as rand_formula = "(1|Species)".

Please let me know if you have any further questions or concerns.

Best regards,
Huang

[aston9090]

Hi Huang,

Thank you for the reply! It solves the error now. May I also ask if that's the right way to test whether there are significant differences across the groups, given that the groups contain different species?

Thanks again!

Best regards
Aston

[FrederickHuangLin]

I lack some context in your study. Based on the code you provided, you have an random intercept term for species, which means the same species has been measure repeatedly. If this is the case, then your code is correct. Otherwise, specifying fix_formula = "Group + Species", rand_formula = NULL might be more appropriate.