I got a series of archaeological objects defined by two measurements: d(iameter) and h(eight).
Since these objects comes from different sites, I was wondering to see if any significant difference between them exists as far as dimensions are concerned.
After having plotted them in a scatterplot and after having displayed the samples by mean of box-plots (see A pdf), I decided to perform the Kruskall-Wallis test along with the Dunn's post-hoc test (as provided by SigmaPlot 11) (see B pdf).
As espected, there are some significant differences in median values among some groups.
What is surprising me is the difference in result between K-W test and Mann-Whiteny test as far as the difference between two groups is concerned.
In K-W test, there is no difference between Eolie vs Tell Taannek_sw groups, whereas according to M-W test (see C and D pdf) there is a significant difference.
Can anyone give me an hint about this discrepancy?
Data file (.xls) is attached.
Advertise on Talk Stats