DISCOVERING STATISTICS USING SAS-经管之家官网！

DISCOVERING STATISTICS USING SAS
ANDY FIELD
JEREMY MILES
2010
SAGE
753pages
This companion is designed for researchers who wish to perform linear models anal-
ysis using the definitional formulas. It can serve as a supplemental text for students
in a college-level theoretical or applied linear models course using The SAS System
(SAS) for computations. This book can also serve as the principal text for a special
topics course in a graduate program in statistics with linear models theory and appli-
cations courses as prerequisites. It may also serve as a useful reference for statistical
analysts who wish to use the SAS/IML product to work out the formulas for new
experimental methods of data analysis.
A linear models course can be taught from different approaches. Some of these
approaches are more theoretical and focus on computation and derivation of the
linear algebra formulations. Such a course might consist of theorem, proof, theorem,
proof, theorem, proof, theorem, lemma, proof, etc. Some theoretical approaches
focus on geometric interpretations of the linear model using projections. Yet other
approachesaremoreappliedandfocusonthecomputer-aidedappliedanalysisusing
high-level computational procedures such as PROC GLM, MIXED, REG, etc. in
SAS, with little time spent with the analytic formulas.
This companion illustrates the theoretical linear algebra approach to teaching
linear model analysis. This companion does not teach linear models concepts, but
demonstrates how SAS/IML can be used to evaluate numerical linear model prob-
lems. It is assumed that the reader does not have any experience using SAS/IML,
but is familiar with DATA steps and basic PROC steps in the SAS system.
All the SAS examples given in this companion are self-contained, and may be
executed as written, without additional programming. In most cases, there are other
SAS procedures that are more appropriate to use in the analysis of linear models
than the IML procedure. In many cases, the analysis will be performed in both IML
and another, more appropriate SAS procedure. This is done for the following reason:
The IML approach is directed at leaning and applying the linear models formulas.
The more “canned” procedures like REG and GLM do not allow students to see the
connection between formulas and the procedures. However, because REG and GLM
are more efficient and numerically accurate than SAS/IML in many applications of
the linear model, they will also be briefly demonstrated. The canned procedures
(such as REG and GLM) are considered the standard for analysis and are provided
to demonstrate that the results obtained using the IML procedures are the same as
those using the canned procedures. This allows students to go from written analytic
formula in say mixed models analysis to IML implementation of that same analytic
formula to high level analysis using PROC MIXED. Adding the step of PROC IML
implementation provides the missing link in the learning process.
Thetopicsselectedforthiscompanion arethetopicstheauthorfoundtobeuseful
in the academic learning of the analysis of linear models. There are many additional
topics useful in the study of linear models that were not included. However, this
should give the reader a complete treatise for a course on the subject.
Exercises have been included to aid the learner. Some exercises were developed
by withholding IML implementation steps for the reader to work through. Other
exercises are developed for perhaps repetitive application of concepts introduced in
the discussion.
The examples presented take advantage of the most recent advances in SAS/IML
and are current as of SAS version 9.2. However, most of the examples will work in
earlier versions of SAS.