Volatility/stability analysis related statistics

Hi there. I am working with longitudinal data and I am trying to see if there is a difference between how variable/volatile the microbiome is over time and between my two groups of interest. I did pairwise differences, but this only uses two time points and I have more data than that.

So, I also looked at the volatility. I have the visualizer, but haven’t been able to figure out which stats to run. I see the notation “accompanying statistical tests may be added in future releases”. Can you point me to any related tests? If you are not sure don’t do extra research, but thought you all might know. If it will be in the next release, then I can wait. :smiley:

Thank you!

1 Like

Hi @Jennifer_Fouquier,
The most appropriate accompaniment for the volatility plots would be linear-mixed-effects (LME); and I believe that method fits what you are describing.

We now have a preprint for q2-longitudinal that provides more extensive examples and description, particularly of volatility plots and LME.

A particularly useful method for you (based on your description) would be to run first-distances on a distance matrix to examine how much each individual changes between sequential sampling times (or from a single time point if you set the baseline parameter), then run LME on that output to examine how different factors impact how individuals’ microbiota change over time. The preprint contains a detailed example of this.

LME can be a but complicated to use and interpret, but a skilled user can extract lots of information out of these models — I would strongly recommend consulting with a statistician if you have not used LME before.

The stats that we don’t have in place yet in the volatility action is a method to determine whether one group displays greater variance over time than another group. This is a much more complicated procedure — but that does not sound like what you are looking for, anyway.

I hope that helps!

3 Likes

Great! Thanks for all the information. I’ll definitely read the paper! I think it’d help if there is a brief cross-reference about these other methods within the volatility visualizer description. When I read the statement about future statistical tests there, I went looking for other resources. Thanks so much!

This topic was automatically closed 31 days after the last reply. New replies are no longer allowed.