Statistical modelling

Another way to compare models is to use what is known as ´deviance´. In simple terms this gives an overall measure of how well the model fits the data. It is calculated as − 2 times what is known as the log likelihood.

Whilst the method of least squares is the method usually adopted when fitting models involving fixed effects only, the method of maximum likelihood is the method used by REML. This method calculates an expression known as the likelihood, which measures how well the model fits the data. The better the fit to the data the smaller is the value of the −2 log likelihood. By comparing the deviance values derived from separate models one can determine which model provides a better fit to the data.