Predicting from the fitted model in MLwiN
To make predictions from a fitted model there is a window with a representation of the model and the user selects just those elements (fixed coefficients, random effects) that they wish included, by selecting the symbols in the lower window. The predicted values, together with standard errors if required are then output ready for plotting etc.
If a random effect is included then the required estimated residual is incorporated.
A typical screen looks as follows:
The screen shows the names instead of symbols and we see that the prediction is for each level 2 unit, incorporating the estimated effect u0j for each unit.