what to do after a chi² test

#1
hey,
i compared the frequency of samples containing zero pupa vs samples containing at least one pupa in different stations. i carried out a chi² test which shows a significant difference. but i wonder how can i show which stations are different than the other? what test could i use?

thanks in advance for your help

Adrine
 

Tart

New Member
#2
Well there are several multiple comparisons tests. Kruskall-Wallis comes to my mind. This test assumes that you have k random samples (possible of different size) from different populations.
H_0 All of the k population distributions are identical
H_1 At least one of the populations tends to yield larger observations than at least on of the populations

If null is rejected then you can do multiple comparisons. It can also be extended to contingency tables.

See W.J.Conover, (1980), "Practical Nonparamentric Statistics", 2nd ed. pp.229-232 (this is edition I have).

Note if you use R - default kruskal.test does not perform multiple comparisons, but it is relatively simply to do.
 

Tart

New Member
#3
by the way following Conover I wrote R functions that perform KW test and perform multiple comparisons, with one version for contingency table. If people are interested I can post them.
 
#4
hey tart,
thanks for your help, i appreciate but i have some more questions
could i compare the station two by two using the chi² test?
because i already know that the null hypothesis is rejected thanks to the chi² test so i could do multiple comparison. does the kruskal wallis test do that multiple comparison?

thanks again