Mann Whitney help

#1
Hi,

I did a small experiment and I had 4 distribuitons.
these are not normal distribuitions, that is why my professor told me to run a
Man Whitney test. She said something about that I need to reject one of the 3 hypothesis that are involved in this field of research.

I have no idea what is the purpose of this test, can anybody help me? please?
all of the results gave me a non significant p values.

thanks!
 

Miner

TS Contributor
#2
The Mann-Whitney test is a nonparametric equivalent to the 2-sample t Test. It tests the hypothesis that the medians of two distributions are equal. The assumptions of the test are that the distributions are the same shape and have equal variances.

You did not provide enough information on your experimental design to tell whether this is the best test. There are also the Kruskal Wallis and Moods Median tests, which are nonparametric equivalents to 1-way ANOVA.
 
#3
I tested a new theory in the absolute adjectives field, I used the Hebrew word for 'full' which is an absolute adjective. I wanted to check if there is a leap in the distribution people will choose the leap as there cutoff point. In the experiment there were 4 distributions of twelve cards in the shape of a cup were made especially for the purpose of this experiment. The cups in each distribution were exact 12 cm of height. Each of the cups was colored in black ink that played the role of the liquid in this case. The liquid in each cup was measured in advanced according to a plan that I made when constructing the distributions with the leaps built-in. the 3 theories are: 1. people would go for the max and chosse the most full cup which is 12 cm. 2. the averaging theory says that people would chosse their cutoff point above the average of the distributions (6.5). 2. the 3d theory says that if there is a leap in the distribution that would be the cutoff point. my professor told me to use Mann Whitney to compare each of the 4 distibutions and I don't understand which of the 3 theories it can reject.
 

Karabiner

TS Contributor
#4
The Mann-Whitney test is a nonparametric equivalent to the 2-sample t Test.
It is often said so. But if one thinks about it, it cannot be true.
Either it ist used on ordinal variables. Then it cannot be equivalent to
a test comparing means. Or it is used on (rank transformed) interval
scaled variables. Then it cannot be equivalent to a comparison of means,
either.
It tests the hypothesis that the medians of two distributions are equal.
I beg to differ. That comparison is done by the median test. The calculation of
the U test (or the Wilcoxon rank sum test) statistic doesn't refer to medians, and
significant U-tests don't necessarily mean that medians are different (although they
often are).

With kind regards

K.