Extended quasi-variances

Model summaries that use quasi variances provide an economical route to approximate inference and graphical presentation in situations where the estimable parameter combinations are contrasts. We show how the quasi-variance notion can be usefully extended in various ways, in particular to cover situations where arbitrarily scaled or rotated contrasts are of interest. The definition, computation and graphical display of such generalized quasi variance summaries will be discussed, with real-data examples.