We used a correlative approach (BEST Bio-Env procedure) to analyze the importance of environmental drivers on large-scale echinoderm assemblage patterns, but it should be noted that this multivariate approach does not indicate directionality of a particular driver.
What approach did you take?
When we consider multivariate models containing multiple brain regions the normalization approach does not seem to be that important.
Which approach do you favor?
The sample size did not allow a multivariate approach that takes into account the potential confounding factors.
In this work, we did not consider a multivariate approach for the two-step methods although it would be interesting to include this option.
More specifically, it is robust against small sample sizes, uses a multivariate approach instead of a univariate one and does not cancel out significant changes in expression levels with non-significant ones.
Second, the large region of interest-based multivariate approach proposed by the authors (Smith et al., 2013) does not solve the problem we have identified but arguably conceals it.
In this particular application of the multivariate approach, accounting for the correlation between outcomes alone (models 2a and 2b) did not reduce the uncertainty around estimates of intervention effects compared to analysing each outcome separately (model 1).
With all conditions (i.e., stimuli) involving identical processes, it is not surprising that this method did not reveal any significant differences, hence the advantage of using a multivariate approach.
