AHuffmyer
commented
4 months ago Here is what I think we could do:
- Analyze all univariate results with haplotype as a random effect
- Run overall permanova at genus level (response ~ site time genus)
- Run genus specific permanova including haplotype (e.g., POC model: response ~ site time haplotype)
- Keep multivariate trajectory and biplots at genus level
- Proceed with variance partitioning including haplotype as a variable
- Keep symbiont ~ host physiology RDA as is, since its looking at physiological indicators within species
- Proceed with showing ITS2 communities by haplotype and running relative abundance anovas including haplotype as main effect
Our preliminary analysis of haplotype effects indicate that there are differences in host and symbiont physiology by haplotype. These are more pronounced in Porites (but one haplotype only found at ⅔ sites) and are small but significant in Pocillopora. ITS2 communities are different by haplotype.
We will need to re-evaluate whether we can include haplotype in multivariate analyses quantitatively as a main effect or if we need to reanalyzed everything with haplotype as a random effect. I think we should present ITS2 by haplotype and include haplotype in variance explained plots as a main effect. We do not have enough replication of each haplotype at every site to include it in the analyses directly as a main effect.
Food for thought! Post with details here: https://ahuffmyer.github.io/ASH_Putnam_Lab_Notebook/Revising-E5-timeseries-results-restructuring-figures-and-filtering-univariate-data/