Permanova + T-test


I have a question regarding the data analysis listed below. I am trying to run a similar study but I can't truly understand why a PERMANOVA was used. Any help would be greatly appreciated.

In my study, I am comparing prey species composition, abundance, pecking rate, and swimming height in 3 min segments and among sites which predators are present and absent. My thought was to run a paired t-test (comparing 3 min segments from the same video) and a standard t-test in comparing the factors listed above.

Analysis in a similar study. "Multivariate hypotheses about fish assemblages was tested with PERMANOVA (Anderson 2001). In comparing 3 min sections of video we analyzes two-factors in a repeated measures design; shark presence and absence (fixed with two levels) and sites (random with 45 levels), with the interaction term being excluded (Anderson et al. 2008). For each analysis, Bray Curtis distances will be generated from untransformed data.
Prior to analysis of univariate variables, a Quantile-Quantile plot will be used to test the assumption of normality. For the variables with normal distributions, hypotheses will be tested with either a paired sample t-test or a standard t-test depending on whether fish were being compared within 3 min segments or among sites with and without sharks, respectively. If data is not normal, we will use a Wilcoxon signed-rank test (Klages et al. 2014)"

Thank you for your time



Super Moderator
If you are only concerned with a single species, then yout t-test method sounds fine. The other study was analysing community data (i.e. several different species) based bray-curtis similarity coefficient, and they obviously also had some univariate response variable (may species richness or somthing like that), which the t-test applies to.
Last edited: