I’ve filtered samples from my dataset so only some of the metadata categories remain (eg., filtered Category 4 samples so only Categories 1-3 remain). After running OLS regression in gneiss, the resulting OLS visualization artifact still contains all of the categories instead of just the filtered ones. Moreover, the “Regression Coefficients Summary” interactive table shows p-values (and corrected p-values) for those “filtered” categories as 0 for all balances, and the coefficients are very large (eg., 1e18).
It appears that gneiss is using all factor levels in the full (not-filtered) metadata table.
It seems that I’m going to have to filter my metadata table to match the filtered feature table prior to running qiime gneiss ols-regression
. Is there any good way of doing this from the command line (or Jupyter) other than:
- export filtered feature table to biom
- convert biom to OTU table (or summarize samples)
- load metadata table + list of samples into R or python
- filter the metadata table based on the list of samples
- write out the filtered metadata table
This is a lot of steps just to filter the metadata table.