Nonparametric Test for Equality of Intermediate Latent Roots in Non-Normal Distribution

Ushizawa, Kenji; Sato, Yoshiharu; Sugiyama, Takakazu

doi:10.5183/jjscs1988.11.9

Cited by 1 publication

(1 citation statement)

References 5 publications

(2 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The Moses test is free from position parameters for PC-scores distribution, but the researchers suggested that the subgroup size should be large but must not be greater than 10 (Hollander, 1968). The comparison between Moses test with Ansari-Bradley and F tests are made in term of efficiency (Ushizawa & Sato, 1998). The Monte Carlo simulation has been used to compare the reliability of the testing comparison between these three tests.…”

Section: Introductionmentioning

confidence: 99%

Comparative Analysis of Classical and Modified Moses Test

Azman

Ahad

Okwonu

2021

JOSMA

View full text Add to dashboard Cite

Moses test is a nonparametric method to test the equality of two dispersion parameters. The Moses test does not assume equality of location parameters, and this fact gives the test wider applicability. However, this test is inefficient since different people applying the test will obtain different values because of a random process. One sub-division may lead to significant results where another does not. To overcome the problem of uniqueness of the result, this study proposed to modify the random selection of the observation for the subsamples based on the ranking procedure to lead for a unique result for each solution. The original and modified Moses test were tested on the same data set. The finding shows that the result for both tests is similar in terms of decision and conclusion. The analysis revealed that the modified Moses test based on ranking approach has a smaller sum of squared values compared to the original Moses test. Thus, the variability of data for each subsample is decreased as well. Ranking approach can be used as an alternative to replacing the random procedure of selecting observations for subsample to overcome the problem of uniqueness in the test statistic.

show abstract

Section: Introductionmentioning

confidence: 99%