Dissertation Information for Robert Pavur NAME: - Robert Pavur
- (Alias) Robert James Pavur
DEGREE:
- Ph.D.
DISCIPLINE:
- Statistics
SCHOOL:
- Texas Tech University (USA) (1981)
ADVISORS: - None COMMITTEE MEMBERS: - None
MPACT Status: Incomplete - Inspected
Title: UNBIASED F-TESTS FOR FACTORIAL EXPERIMENTS WITH CORRELATED DATA
Abstract: "This research finds, under certain assumptions, the most general form that the correlation matrix can be for the observations in a k-way factorial experiment, so that the usual F test can be corrected by a constant. A method is provided for computing the correction constants for the F tests from a given correlation matrix. This correlation matrix is a variation on the theme of equicorrelation. A set of random variables is equicorrelated if the correlation coefficient is constant for any pair of observations.
The corrected F tests are shown to be the likelihood ratio tests. From this result, it follows that the corrected F is an unbiased uniformly most powerful test among the class of invariant test. An extension is made to the multivariate analysis of variance. A Monte Carlo simulation study is made in which the corrected F statistic on the data and the corrected F statistic on the ranks is examined. These two test statistics are compared under certain situations when not all of the usual assumptions of an ANOVA are present.
The factorial lay-out used in the simulation study is a two-way factorial with interaction and having 5 observations per cell. Cumulative plots for the corrected F statistic and for the corrected F statistic on the ranks are given for this two-way layout using observations from normal, lognormal, uniform, double-exponential, and Cauchy distributions. The corrected F statistic on the ranks appeared in most cases to be as robust as the corrected F statistic and out-performed the corrected F statistic on many of the empirical power figures."
| |
MPACT Scores for Robert Pavur A = 0
C = 4
A+C = 4
T = 0
G = 0
W = 0
TD = 0
TA = 0
calculated 2010-09-28 00:00:57
Advisors and Advisees Graph
|