THESIS IDEA: Currently there are about 50 different tests for normality, and there is no way of deciding which of these are good or bad. The main idea of the thesis is to find ways of comparing tests so that we can decide which tests to choose in practical applications. For more details, see my homepage.
2. In One-Way ANOVA, Normality of the residuals is the underlying assumption. I have calculated the Type_I Error rates for this test and against some particular alternatives, I got type-I errors up to 25% instead of 5%.
My next step is to find the best test against those alternative distributions.
3. Purpose based normality testing
Tests under consideration are:
i. GQ-test
ii. Levene's test of equality of variances
ii. BF-test of equality of variances
iii. ONE-WAY ANOVA (A test of mean comparisons)
Next step is to find the local stringencies of all the test then by comparing the test based on their stringencies, we will selected a best test.
PROGRESS REPORT:
1. Calculating the powers of the fifteen selected tests against the Mixture of two t-distributions.
X=alpha*T1+(1-alpha)*T2
AZ: Tasks to do:
1. Create a page entitled Ph.D. Thesis on your website.
2. Create a subpage Annotated Bibliography. List all references related to your thesis on this page. Here is a internet program meant for managing references. I dont know how it works, and I would like some of students to check it out to see itf it is useful for the rest of us: http://www.mendeley.com/how-it-works/
3. For the main references create a subpage giving an abstract of the main findings and also a link enabling download of the full refernce. You can collect these reference into subgroups and put several on the same page.
4. Some of the key subgoals towards the main goal of your thesis are listed below -- indicate your progress in the left panel.