Browsing by Author "Fekadu, Addisu"

Now showing 1 - 1 of 1
  • No Thumbnail Available
    Item
    Applications of Repeated Measures Analysis on Enset Plant (Ensete Uentricousum )
    (Addis Ababa University, 2008-07) Fekadu, Addisu; Taye, Girma (PhD)
    Repeated measures analyses have become the most interesting areas In psychological, health and agricu ltural researches. Repeated meaSU!'es data are measurements taken several times from the same subject. Such data tend to be serially correlated. Measurements taken close in time are potentially highly correlated than those taken far apart in time. Hence, they require special methods of analysis. In this paper, four major approaches are used to analyze repeated measurements taken from thirty three varieties of ensel plants each measured at four successive time points. The study provides summary statistics, results based on repeated measures analysis of variance (Spli t-plot in time ANOVA), multivariate analysis of variance (MAN OVA), and mixed model methods. Each method is described briefly. In order to apply the repeated measures ANOYA, compound symmetry assumptions of covariance structures should be met. Whether the data fulfils this structure is tested. For those data which do not satisfy this criterion, the degree of freedom is adjusted for F test statistics by Huynh-Feldt (H-F) or Oreenhouse-Oeisser (0-0) epsilons. The multi variate approach is less restrictive but lacks power given that the repeated measures AN OVA assumptions are sati sfied and the sample size is small. In all the methods considered, SAS was used to analyze the data. The results using uni vari ate, multivariate and mixed approaches of repeated measurements of ensel plants show that the main effects of variety and time as well as the interaction effect of veri ety by time were found to be significan t. It was found that mixed model approach provides a very flexibl e environment in which the covariance structure can be modeled. Besides, the mixed model permits selection of the covariance structure that best fits the data at hand and enables to compute efficient estimates of fixed effects and valid standard erro rs of the estimates.