Testing equality of two normal means using a variance pre-test
Since the two-sample t-test presupposes equality of variances, often a preliminary F-test is advised to check this assumption. However, very large sample sizes are necessary before this procedure behaves in a satisfactory manner. In the present note this phenomenon is explained using second-order asymptotics. Simple and accurate expressions are derived relating the sample sizes required to the underlying parameters.
| Year of publication: |
1998
|
|---|---|
| Authors: | Albers, Willem ; Boon, Pieta C. ; Kallenberg, Wilbert C. M. |
| Published in: |
Statistics & Probability Letters. - Elsevier, ISSN 0167-7152. - Vol. 38.1998, 3, p. 221-227
|
| Publisher: |
Elsevier |
| Keywords: | One-sample problem Two-sample problem t-test Preliminary test Asymptotics |
Saved in:
Saved in favorites
Similar items by person
-
Normal control charts with nonparametric safeguard
Albers, Willem, (2009)
-
Albers, Willem, (2006)
-
Accurate test limits under prescribed consumer risk
Albers, Willem, (1997)
- More ...