Replicated samples in divergence functions

[TuomasBorman]

getStepwiseDivergence:

I think it was not optimal to expect user to subset data so that there were no replicated samples. Now user gets warning if there are replicated timepoints, and the average is calculated.

[TuomasBorman]

Unit tests are still needed.

[TuomasBorman]

Now, if there are repeated samples, i.e.,

• getBaselineDivergence: there are multiple samples in first time point for single group
• getStepwiseDivergence: there are multiple samples in single time point for single group

the repeated samples are handled by calculating average. The output is like this:

> res <- getStepwiseDivergence(
+     tse, time.col = "time", group = "group", method = "euclidean") 
Warning message:
Some samples are associated with multiple reference samples. In these cases, the reference time point includes multiple samples, and their average is used. 
> res
DataFrame with 20 rows and 3 columns
         divergence time_diff                   ref_samples
          <numeric> <numeric>                        <list>
sample1      189737         2               sample5,sample9
sample2      379473         5                      sample14
sample3      326769         3 sample1,sample11,sample13,...
sample4      202386        94  sample2,sample8,sample10,...
sample5          NA        NA                            NA
...             ...       ...                           ...
sample16    63245.6         5                      sample14
sample17   316227.8         2               sample5,sample9
sample18   265631.3        94  sample2,sample8,sample10,...
sample19   379473.3         2               sample5,sample9
sample20   328876.9        94  sample2,sample8,sample10,...

[TuomasBorman]

#69 (comment)

Hmm sometimes also other variables might be necessary to consider in addition subject and time. For instance, subject + time + bodysite.

I think the simplest solution for this is to create new grouping variables to colData.

tse[["group_bodysite"]] <- paste0(tse[["group"]], tse[["bodysite"]])
res <- getBaselineDivergence(tse, group = "group_bodysite")

Also, one can always input own reference samples.

[TuomasBorman]

"Multiple reference samples" handling is now moved to mia::getDivergence, it is more straightforward solution: microbiome/mia#680