[Solved] Use mean or not?

[smir]

Hi everbody,

Here's probably a quite simple question but I just want to be sure before doing tests on it.

I'm not sure if I have to take the means of my data or not, here's a short description of my experiment:

I have 18 single blocks where I measured, at each block on two areas. So for each block I have 2 measurements (this repeated over a longer time span).

So should I take the mean of these two measurements? If not, wouldn't it be pseudoreplication because they belong together?

I hope the information is enough for an answer, otherwise I'll explain the experiment better.

Thanks in advance!

[Dason]

What type of analysis do you plan on running? You could take means as long as you have the same number of pseudoreplicates in each group. Otherwise you could include all of the data and model it as a subsampling problem by setting up an acceptable correlation structure.

[smir]

I'm planning to run a Kruskal-Wallis-test.
I do have the same number of pseudoreplicates in each group.

So I'll rather take the mean.

Thanks!