Analysis of unbalanced mixed model data: A case study comparison of ANOVA versus REML/GLS

摘要: Major transition has occurred in recent years statistical methods for analysis of linear mixed model data from variance (ANOVA) to likelihood-based methods. Prior the early 1990s, most applications used some version because computer software was either not available or easy use ANOVA is based on ordinary least squares computations, with adoptions models. Computer programs such methodology were plagued technical problems estimability, weighting, and handling missing data. Likelihood-based mainly a combination residual maximum likelihood (REML) estimation covariance parameters generalized (GLS) mean parameters. Software REML/GLS became readily but still universally embraced. Although many computational inadequacies have been overcome, conceptual remain. Also, emerged, as need adjustments effects due estimating This article attempts identify major ANOVA, describe which remain REML/GLS, discuss new REML/GLS.