THE ROBUSTNESS OF THE STUDENT t TEST WHEN SAMPLING FROM A WEIBULL DISTRIBUTION.
NAVAL POSTGRADUATE SCHOOL MONTEREY CALIF
Pagination or Media Count:
When testing with the t-test, it is assumed that the sample under investigation is from a normal population. The purpose of this thesis is to examine the sensitivity of the t-test to violations of this normality assumption. A computer simulation was performed to draw sets of 10,000 samples from an infinite Weibull population. A t-test was performed on each sample to test the null hypothesis H sub 0 mu or mu sub 0 where mu sub 0 was the true mean of the Weibull population. The number of times that H sub 0 was rejected was recorded for all combinations of eight levels of significance, samples ranging in size from 2 to 31, and for values of the parameters of the distribution lambda 1,2,3 and beta 1,2,3. Author
- Statistics and Probability