Title: A nonparametric test of 2-factor factorial designs with unequal replicates for multivariate data
Authors: Marcus Jude San Pedro - School Statistics, University of the Philippines Diliman (Philippines) [presenting]
Mara Sherlin Talento - College of Arts and Sciences, University of the Philippines Los Banos (Philippines)
Erniel Barrios - University of the Philippines (Philippines)
Abstract: MANOVA requires equal replicates for treatment with factorial design for its robustness to normality assumption and orthogonality of the basis of vector space. However, when data set is composed of Poisson and Normal error terms and replicates are unequal, normality and sphericity assumption will be rejected. The aim is to propose a nonparametric approach in testing interaction effect of two-factor factorial design even for highly unbalanced replicates and multivariate data with Poisson random variable. Results have shown small probability of committing type 1 error and increasing power for increasing effect of second treatment, increasing coefficient of interaction effect and increasing mean response of levels of second treatment effect, holding other factors constant.