This paper develops maximum likelihood estimates for jointly modelling the mean and covariance matrix, for unbalanced repeated measures, using an unconstrained parametrization.
Marginal models do not require any distributional assumptions beyond correctly modelling the mean (average) outcome, yet when combined with robust standard errors they can provide appropriate inferences.
Instead of modelling the mean of the influenza incidence rates series, it has been discussed in [ 10] that it is more appropriate to model the first-order differenced series (formed by the differences between rates in consecutive weeks).
It should also be stressed that we assumed the mean dose used during the clinical trial, while it could be argued that after the first 8 weeks of modelling the mean final dose should be used.
The variable-dispersion beta regression model is a type of double-index regression model [ 13], as it contains two regression equations, one modelling the mean as a function of covariates and the other modelling the dispersion as a function of covariates.
To avoid a scenario where an effect of a covariate on the level of SRH at diagnosis was misinterpreted as an effect on the estimated change between the two time points, the covariates were also used for modelling the mean SRH at diagnosis.
