The lack of fit source of variation associated with a fitted linear model is analyzed with respect to a specified alternative model. A partitioning of the lack of fit sum of squares is carried out in order to isolate and identify a complete set of causes, based on an alternative model, which are thought to contribute to the lack of fit variation. These causes are expressed as linear functions of the parameters of the alternative model which were not included in the fitted model. Three numerical examples are presented to illustrate the procedure.
展开▼