Quiz #10
The answer is given in the answer box below.
Scroll down for an explantion for the answer (given in red).
Be sure to put your name on this quiz. Write the letter corresponding
to the correct answer in the answer box below.
It is common to use a chi-square distribution as an approximate null sampling distribution for the Kruskal-Wallis
test statistic because tables and software needed to obtain exact p-values are hard to find, and for many
situations there is no good way to obtain an exact p-value. Fortunately, the approximation works fairly well unless
the sample sizes are too small. If one is comparing four distributions with the Kruskal-Wallis test, and one is
going to use equal sample sizes, what is the smallest number of observations per sample that can be used to perform a
reasonably accurate (approximate) size 0.05 test? (Make use of the rule-of-thumb that I mentioned in last week's
lecture.)
- [ A ] 3
- [ B ] 5
- [ C ] 10
- [ D ] 12
- [ E ] 25
- [ F ] 30
- [ G ] 50
In order to do a decent (with regard to validity) approximate size 0.05 test using the (routinely relied upon)
chi-square approximation with the Kruskal-Wallis test, one should have 6 observations per sample if there are just
three samples, and one should have at least 5 observations per sample if there are four or more samples.
So the correct answer is [ B ]. (Given that we have little choice (except not to use the test) but to rely on the
chi-square approximation for the K-W test, and given that SPSS will compute the p-value for us, one of the important
pieces of information pertaining to this test that you should know is when you can trust the approximate p-value.)