Comparison of non-parametric tests of ordered alternatives for repeated measures in randomized blocks

AKDUR H. T. K.

COMMUNICATIONS IN STATISTICS-SIMULATION AND COMPUTATION, cilt.51, sa.7, ss.4146-4158, 2022 (SCI-Expanded) identifier identifier

  • Yayın Türü: Makale / Tam Makale
  • Cilt numarası: 51 Sayı: 7
  • Basım Tarihi: 2022
  • Doi Numarası: 10.1080/03610918.2020.1740262
  • Dergi Adı: COMMUNICATIONS IN STATISTICS-SIMULATION AND COMPUTATION
  • Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Academic Search Premier, Applied Science & Technology Source, Business Source Elite, Business Source Premier, CAB Abstracts, Compendex, Computer & Applied Sciences, Veterinary Science Database, zbMATH, Civil Engineering Abstracts
  • Sayfa Sayıları: ss.4146-4158
  • Anahtar Kelimeler: Circular bootstrap, Non-parametric trend tests, Randomized blocks, Repeated measures, RANK-BASED TESTS, HYPOTHESES, TREND
  • Gazi Üniversitesi Adresli: Evet

Özet

Page, Jonckheere, Hollander tests are the most common non-parametric tests for ordered alternative pattern in randomized blocks where contain independent observations. However, when randomized blocks include correlated observations, properties of null distribution for these tests under independence cannot be applied to the situations where randomized blocks contain dependent observations. Circular bootstrap versions of these tests are suggested to test for ordered alternative hypothesis in randomized blocks which include correlated measurements when the underlying distributions are not known. In this study the effect of correlation coefficient, number of treatment/time point and block sizes on type 1 error and power values of some existing non-parametric test statistics based on the circular bootstrap method are taken into consideration with a stationary auto regressive process. The significance levels and power values of these tests based on circular bootstrap method are simulated by using the statistical software package R, 3.4.2. Based on findings in the simulation study, circular bootstrap version of original Page Test provides better power values among other trend statistics. Finally, these circular bootstrap-based non-parametric tests are applied to a real life dataset.