I have another question that I would like to ask you. I am working on a data set made up by 2 groups (corresponding to treatment vs control), of 5 individuals each, and all of these individuals have been measured at 5 time points, i.e. in a longitudinal way. I have seen that the ‘multilevel’ options available with Mixomics allow the study of longitudinal series, but I am not sure if, beyond the comparison of longitudinal data, it is also possible to take into account the comparison between two groups or treatments.
Beyond that, I was wondering which could be the guidelines to follow to select the proper statistical method to analyze the data. Is there a specifically adapted approach when working on longitudinal series?
Or is it possible to ‘freely’ select one methodology or another one?
Thanks in advance
For 5 time points, or more, I recommend you use a spline approach, combined with PLS.
You can read this paper which highlights the type of analysis you could do with longitudinal data (ignore the microbiome aspect):
Statistical challenges in longitudinal microbiome data analysis
We have developed a package for this also (I’m trying to get the recording out and I’ll post it on our website):
A generic multivariate framework for the integration of microbiome longitudinal studies with other data types
→ A follow up of Straube et al.
timeOmics: an R package for longitudinal multi-omics data integration
→ And the associated package
Interpretation of network-based integration from multi-omics longitudinal data
→ And a follow up of the follow up!
DynOmics to identify delays and co-expression patterns across time course experiments
→ This one could also be interesting if you are interested in detecting time delays
Thanks for your answer, the approaches that you suggest seem promising, and I would like to try at least one of them in the next days. Anyway, on my side there is another issue now. I have just realized that I won’t be able to use the last time point of mine because of quality issues.
Therefore, now I have just 4 time points available. Do you think I could still use the methods you indicate, or would it be better to try something simpler, i.e. like DA analysis?
Best regards and thanks in advance,
You can still use 4 time points (I think) for the lmms / timeOmics.
I think starting first with a PCA, and then a PCA multilevel will already give you some insights about the longitudinal aspect of your data (i.e if there is a time trend).
Thanks again for your answer. I have already performed an sPCA, which indeed shows that the data are arranged longitudinally. I will also run a multilevel PCA, as you suggest, and then I will try ‘timeomics’.
Have a nice day and thanks again,