This edition has been completely rewritten. The main features of the new edition are the following:
1. The mathematical, statistical threshold has been lowered in order to make the material more accessible to students with only an elementary prior knowledge of statistics.
2. All the inference procedures for general linear model have been derived as special cases of a single basic procedure, namely, the testing of a set of linear restrictions on the parameters of the model (Chapter 5).
3. Every effort has been made to cover both new and old topics on which substantial for work has been done in recent years and which are thought to be significant and enduring rather than passing fancies.
4. The problem sets have been extended to become truly Anglo-American, with offerings from the Royal Statistical Society plus the universities of Cambridge, London, Manchester, and Oxford on one side of the pond, and Chicago Michigan, Yale, and Washington on the other side.