help with correct adonis model specification

[ramiroricardo]

Dear all,

I am analysing a microbiome dataset in which mice were sampled at two time points, before and after a treatment. I want to test if the communities are different before and after treatment.

I would like to use mouse_id as a blocking factor, as the initial state of the microbiomes differs between mice.

I understand that adonis is not designed to deal with random effects, but I was wondering if this could be dealt with by fitting mouse_id as a fixed effect? (as one could do in a linear model)

I get very different results when running the models with and without mouse_id as a fixed effect (here I am using a Bray-Curtis distance matrix).

Model without mouse_id

adonis(ps_recipient_KOWT_bc~as.factor(timepoint),data=sample_df_KOWT)

'nperm' >= set of all permutations: complete enumeration.
Set of permutations < 'minperm'. Generating entire set.

Call:
adonis(formula = ps_recipient_KOWT_bc ~ as.factor(timepoint),      data = sample_df_KOWT) 

Permutation: free
Number of permutations: 719

Terms added sequentially (first to last)

                     Df SumsOfSqs MeanSqs F.Model      R2 Pr(>F)
as.factor(timepoint)  1   0.52713 0.52713  17.916 0.81749    0.1
Residuals             4   0.11769 0.02942         0.18251       
Total                 5   0.64482                 1.00000  

Model with mouse_id fitted as a fixed effect

adonis(ps_recipient_KOWT_bc~as.factor(timepoint)+mouse_id,data=sample_df_KOWT)

'nperm' >= set of all permutations: complete enumeration.
Set of permutations < 'minperm'. Generating entire set.

Call:
adonis(formula = ps_recipient_KOWT_bc ~ as.factor(timepoint) +      mouse_id, data = sample_df_KOWT) 

Permutation: free
Number of permutations: 719

Terms added sequentially (first to last)

                     Df SumsOfSqs MeanSqs F.Model      R2  Pr(>F)  
as.factor(timepoint)  1   0.52713 0.52713 18.9481 0.81749 0.03333 *
mouse_id              2   0.06205 0.03102  1.1152 0.09623 0.63333  
Residuals             2   0.05564 0.02782         0.08629          
Total                 5   0.64482                 1.00000          
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Would the latter model be correct?

I have also thought of using strata=mouse_id, though since I only have two samples per mouse, I think this would lead to a rather low number of permutations.

Kind regards,

Ricardo Ramiro