Added start of LOCOM analysis code

[nearinj]

Added LOCOM code and fixed a small bug in the momspi setup.

Note that LOCOM doesn't allow for models that don't have covariates (which is odd to me...). I set a dumb covariate where all values are the exact same. Happy to do something else as advised....

Here is a similar figure comparing the observed and calibrated LFC between low and high diversity samples. Not sure why in some cases they do not match up exactly with the original supplemental figure 10.

Further as expected LOCOM expects raw read counts so no normalization was used in both the observed and calibrated data.

[mikemc]

Thanks @nearinj !

Note that LOCOM doesn't allow for models that don't have covariates (which is odd to me...). I set a dumb covariate where all values are the exact same.

Interesting; you mean that you have to set C to something in the locom() function? In that case, can you try setting C=0 in locom() or creating a dummy variable that is all 0?

Here is a similar figure comparing the observed and calibrated LFC between low and high diversity samples. Not sure why in some cases they do not match up exactly with the original supplemental figure 10.

This is very useful! Overall the results look similar to our figure 10, but with an overall shift downwards which I think makes sense given the expected difference in looking at LFC in proportion versus a ratio.

Do you know why the taxa in your figure are different than those in SI Fig 10? e.g. your figure has Streptococcus spp. and SI Fig 10 doesn't)

It seems useful to understand why there is still a small difference with and without calibration - possibly this is being affected by the zero-replacement method adjust_dirichlet(). But good to see that the difference is quite small, as expected.