30.04.2013 Views

SAS/STAT 9.2 User's Guide: The MIXED Procedure (Book Excerpt)

SAS/STAT 9.2 User's Guide: The MIXED Procedure (Book Excerpt)

SAS/STAT 9.2 User's Guide: The MIXED Procedure (Book Excerpt)

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

3940 ✦ Chapter 56: <strong>The</strong> <strong>MIXED</strong> <strong>Procedure</strong><br />

In selecting the base distribution, PROC <strong>MIXED</strong> makes use of the fact that the fixed-effects parameters<br />

can be analytically integrated out of the joint posterior, leaving the marginal posterior density<br />

of the variance components. In order to better approximate the marginal posterior density of the<br />

variance components, PROC <strong>MIXED</strong> transforms them by using the MIVQUE(0) equations. You<br />

can display the selected transformation with the PTRANS option or specify your own with the<br />

TDATA= option. <strong>The</strong> density of the transformed parameters is then approximated by a product of<br />

inverted gamma densities (see Gelfand et al. 1990).<br />

To determine the parameters for the inverted gamma densities, PROC <strong>MIXED</strong> evaluates the logarithm<br />

of the posterior density over a grid of points in each of the transformed parameters, and you<br />

can display the results of this search with the PSEARCH option. PROC <strong>MIXED</strong> then performs a<br />

linear regression of these values on the logarithm of the inverted gamma density. <strong>The</strong> resulting base<br />

densities are displayed in the “Base Densities” table; for ODS purposes, the name of this table is<br />

“BaseDen.” You can input different base densities with the BDATA= option.<br />

At the end of the sampling, the “Acceptance Rates” table displays the acceptance rate computed<br />

as the number of accepted samples divided by the total number of samples generated. For ODS<br />

purposes, the name of the “Acceptance Rates” table is “AcceptanceRates.”<br />

<strong>The</strong> OUT= option specifies the output data set containing the posterior sample. PROC <strong>MIXED</strong> automatically<br />

includes all variance component parameters in this data set (labeled COVP1–COVPn),<br />

the Type 3 F statistics constructed as in Ghosh (1992) discussing Schervish (1992) (labeled T3Fn),<br />

the log values of the posterior (labeled LOGF), the log of the base sampling density (labeled<br />

LOGG), and the log of their ratio (labeled LOGRATIO). If you specify the SOLUTION option<br />

in the MODEL statement, the data set also contains a random sample from the posterior density<br />

of the fixed-effects parameters (labeled BETAn); and if you specify the SOLUTION option in<br />

the RANDOM statement, the table contains a random sample from the posterior density of the<br />

random-effects parameters (labeled GAMn). PROC <strong>MIXED</strong> also generates additional variables<br />

corresponding to any CONTRAST, ESTIMATE, or LSMEANS statement that you specify.<br />

Subsequently, you can use <strong>SAS</strong>/INSIGHT or the UNIVARIATE, CAPABILITY, or KDE procedure<br />

to analyze the posterior sample.<br />

<strong>The</strong> prior density of the variance components is, by default, a noninformative version of Jeffreys’<br />

prior (Box and Tiao 1973). You can also specify informative priors with the DATA= option or a<br />

flat (equal to 1) prior for the variance components. <strong>The</strong> prior density of the fixed-effects parameters<br />

is assumed to be flat (equal to 1), and the resulting posterior is conditionally multivariate normal<br />

(conditioning on the variance component parameters) with mean .X 0 V 1 X/ X 0 V 1 y and variance<br />

.X 0 V 1 X/ .<br />

<strong>The</strong> distribution argument in the PRIOR statement determines the prior density for the variance<br />

component parameters of your mixed model. Valid values are as follows.<br />

DATA=<br />

enables you to input the prior densities of the variance components used by the sampling<br />

algorithm. This data set must contain the Type and Parm1–Parmn variables, where n is the<br />

largest number of parameters among each of the base densities. <strong>The</strong> format of the DATA=<br />

data set matches that created by PROC <strong>MIXED</strong> in the “Base Densities” table, so you can<br />

output the densities from one run and use them as input for a subsequent run.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!