Chapter 49: The NESTED Procedure


Overview

The NESTED procedure performs random effects analysis of variance for data from an experiment with a nested (hierarchical) structure. [ *] A random effects model for data from a completely nested design with two factors has the following form:

click to expand

where

y ijr

is the value of the dependent variable observed at the r th replication with the first factor at its i th level and the second factor at its j th level.

¼

is the overall (fixed) mean of the sampling population.

± i , ² ij , ijr

are mutually uncorrelated random effects with zero means and respective variances , , and (the variance components ).

This model is appropriate for an experiment with a multi-stage nested sampling design. An example of this is given in Example 49.1 on page 2994, where four turnip plants are randomly chosen (the first factor), then three leaves are randomly chosen from each plant (the second factor nested within the first), and then two samples are taken from each leaf (the different replications at fixed levels of the two factors).

Note that PROC NESTED is appropriate for models with only classification effects; it does not handle models that contain continuous covariates. For random effects models with covariates, use either the GLM or MIXED procedure.

Contrasted with Other SAS Procedures

The NESTED procedure performs a computationally efficient analysis of variance for data with a nested design, estimating the different components of variance and also testing for their significance if the design is balanced (see the Unbalanced Data section on page 2990). Although other procedures (such as GLM and MIXED) provide similar analyses, PROC NESTED is both easier to use and computationally more efficient for this special type of design. This is especially true when the design involves a large number of factors, levels, or observations.

For example, to specify a four-factor completely nested design in the GLM procedure, you use the form

  class a b c d;   model y=a b(a) c(a b) d(a b c);  

However, to specify the same design in PROC NESTED, you simply use the form

  class a b c d;   var y;  

In addition, other procedures require TEST statements to perform appropriate tests, whereas the NESTED procedure produces the appropriate tests automatically. However, PROC NESTED makes one assumption about the input data that the other procedures do not: PROC NESTED assumes that the input data set is sorted by the classification (CLASS) variables defining the effects. If you use PROC NESTED on data that is not sorted by the CLASS variables, then the results may not be valid.

[ *] PROC NESTED is modeled after the General Purpose Nested Analysis of Variance program of the Dairy Cattle Research Branch of the United States Department of Agriculture. That program was originally written by M.R. Swanson, Statistical Reporting Service, United States Department of Agriculture.




SAS.STAT 9.1 Users Guide (Vol. 4)
SAS.STAT 9.1 Users Guide (Vol. 4)
ISBN: N/A
EAN: N/A
Year: 2004
Pages: 91

flylib.com © 2008-2017.
If you may any questions please contact us: flylib@qtcs.net