We employed three related analyses: a least squares mixed-model ANOVA, a maximum likelihood mixed-effects analysis and a Bayesian multilevel (or hierarchical) model using Gibbs Sampling (MCMC).
Maximum likelihood mixed-effects models also have several other advantages over the classical least squares approach regarding better handling of missing values, parameter estimation and prediction [9].
We mention this because one advantage of maximum likelihood mixed-effects models is that there appears to be greater agreement regarding statistical testing [24].
We have already seen that, despite some differences in the details, the least squares ANOVA and maximum likelihood mixed-effects model are qualitatively in agreement.
We used maximum likelihood estimation with repeated measures in SAS Proc Mixed to account for the correlations over time, with an unstructured covariance matrix.
All regression models used maximum likelihood estimates and we controlled for autocorrelation by correcting for autoregressive effects.
Estimates used maximum likelihood.
All analyses were conducted using maximum likelihood, random-effects models weighted by the inverse of the variance.
Outcomes (change in serum pH, pCO2, HCO3, and base excess) used a maximum-likelihood, mixed-effects repeated-measures model (MMRM) with all longitudinal observations.
We used linear mixed-effects models fit using maximum likelihood to investigate the relationship between relative host use and host reproductive biology.
