ABSTRACT
It is a common experience in scientific research to observe more than one response (output) by modifying one or more input conditions (interventions). Study on the effect of ultra-sound treatment of seeds to plant height, shoot length, flowering pattern etc., is an example with multivariate outcomes.
This chapter explains Multivariate Analysis of Variance (MANOVA) a procedure to test whether the mean values of a panel of variables differ significantly across several groups of subjects. Unlike the univariate ANOVA here the outcome (response) has to take into account covariance matrices among the groups. The role of Wilk’s lambda criterion and the Box’s M-test in understanding data are discussed. A stepwise method in SPSS is also provided (with real data) to assist the practitioner. The utility of estimated marginal means and their standard errors is also highlighted. Inclusion of continuous covariates on the mean vector of outcomes is also discussed. Simple predictive models are explained with an Excel template. (157)
