Abstract:
A statistical problem arises in estimating a variance component in an analysis of variance in various experimental designs when the estimate fo the variance component becomes negative. Various authors [Hartely (1950), Link (1950), Patnaik (1950) Leslie and brown (1966), Barlow et al. (1969), Pearson (1932, 1966) and Ahmad (1974, 1978)] have devised altemative methods to estimate variance that guarantee non-negative estimate of variance components. However, the distribution of the ratio of two ranges is restricted to two samples of unequal sizes. In this paper, an attempt has been made to extend Leslie and Brown (1966) F-test (FLB) based on k samples of unequal sizes. The sampling distribution of ranges is derived based on k-samples of unequal sizes from a standard normal population. Since the integration is difficult to compute the critical regions, Monte Carlo procedure has been adopted. An example is given to compare it with Snedecor F-test (FS) and FLB.
Page(s):
317-321
DOI:
DOI not available
Published:
Journal: Pakistan Journal of Statistics, Volume: 25, Issue: 3, Year: 2009