Best Statistica Test to compare multiple means.

[Tal Litmanovitch]

Hi there.

For my project I am comparing measured concentrations of phosphorus across 4 different locations in a stormwater pond. Water flows in through the Inflow, moves into the Forebay,, then Midbay and exits at the Outflow. My main goal is to compare significant differences in mean concentrations between the 4 locations (most importantly Inflow vsOutflow, but additionally between any of the other locations). Also I should state my data is non-parametric.

As my question states, I am looking for the best test to compare whether these means are statistically different from eachother to determine whether the stormwater pond is causing an increase or decrease in phosphorus concentration between the locations before it exports water. I have read about the Kruskal Wallis/Dunn testbut as I understand it does not compare the means (averages) but the mean ranks based on medians instead, which is not of much use. I have also tried the Mann-Whitney U test but I've heard that its results aren't trustworthy if you are comparing two groups of data (e.g 2 locations) when there are more than two groups (in my case 4 locations). Would the Boniferroni post hoc test be valid Or perhaps some other test that is more robust

 

Thank you!