Model

PROC MIXED; CLASS centre treat; MODEL dbp = dbp1 treat/ SOLUTION; RANDOM centre;

PRIOR / NSAMPLE=11000 OUT=Samples;

Model Information

Data Set

Dependent Variable Covariance Structure Estimation Method Residual Variance Method Fixed Effects SE Method Degrees of Freedom Method

WORK.A dbp

Variance Components

REML

Profile

Model-Based

Containment

Class Level Information

Class

Levels

Values centre

32 35 36 37 40 41 ABC

treat

Dimensions

Covariance Parameters Columns in X Columns in Z Subjects

Max Obs Per Subject

Number of Observations

Number of Observations Read Number of Observations Used Number of Observations Not Used

288 288 0

Iteration History Iteration Evaluations -2 Res Log Like Criterion

0

1

2072.

.30225900

1

3

2056.

. 50198964

0.

00003184

2

1

2056.

.47608790

0.

00000030

3

1

2056.

.47585363

0.

00000000

Convergence criteria met.

Covariance Parameter Estimates

Cov Parm Estimate centre 7.8248

Residual 70.9263

Fit Statistics

-2 Res Log Likelihood

2056.

5

AIC (smaller is better)

2060.

5

AICC (smaller is better)

2060.

5

BIC (smaller is better)

2063.

2

Solution for Fixed Effects

Standard

Effect

treat

Estimate

Error

DF

t Value

Pr

> |t|

Intercept

60.5169

11

.1119

28

5.

45

<

0001

dbp1

0.2800

0

.1071

256

2.

61

0.

0095

treat

A

2.9807

1

.2108

256

2.

46

0.

0145

treat

B

1.9474

1

.2387

256

1.

57

0.

1171

treat

C

0

Type 3 Tests of Fixed Effects Num Den

Effect DF DF F Value Pr > F

dbp1 1 256 6.84 0.0095

treat 2 256 3.10 0.0466

Posterior Sampling Information

Prior Algorithm Sample Size Seed

Jeffreys

Independence Chain

11000

502406001

Base Densities

Density Type

Parm1

Parm2

ig ig

9.0027 132.84

1899.1 9395.7

Acceptance Rates

Boundary Constraints 1.0

Sampling

0.98

Thus, most of the output is identical to that obtained from the default REML analysis. Note that the parameter estimates produced do not relate to the Bayesian estimates. These will need to be obtained from the dataset created by the PRIOR statement 'samples', which contains the sampled parameters (see below). At the end of the usual PROC MIXED output information is provided on: the posterior sampling information used for the Bayesian analysis; the inverse gamma distribution parameters used for the base density function; and the rate of samples accepted in the rejection sampling. However, usually it is not necessary to use this information.

0 0

Post a comment