Nonparametric Test of Interaction Effect for 2²-Factorial Design with Unequal Replicates: Case of Poisson-Normal Multivariate Data

Year: 2019       Vol.: 68       No.: 2      

Authors: Mara Sherlin D. Talento, Marcus Jude P. San Pedro and Erniel B. Barrios

Abstract:

Multivariate Analysis of Variance (MANOVA) is fairly robust to the normality and constant variance assumptions provided that the data is generated from a balanced design. Issues with hypothesistesting arises when error-distribution is non-normal or when the data is generated from an unbalanced design. We propose a nonparametric method of testing interaction effect of the twofactor factorial design with multivariate response and possibly highly unbalanced replicates. Simulation studies indicated that the test is correctly-sized and increasing power with increasing effect-size, and increasing sample size. The parametric test based on MANOVA is incorrectl

Keywords: two-factor factorial design, unbalanced replicates, nonparametric, multivariate data, Poisson and Normal data

Download this article: