INFERENCE Subcommand (META REGRESSION command)
The META REGRESSION command calls the meta-regression procedure. The optional INFERENCE subcommand specifies the estimation methods.
- MODEL
- The optional keyword specifies the model type.
- RANDOM
- The default setting that builds the random-effects model.
- FIXED
- Assumes the fixed-effects model, but does not include the dispersion
parameter. When
FIXED(MULTIPLICATIVE)
is specified, it introduces the multiplicative dispersion parameter to the weights and the variance of the error terms.
- INTERCEPT
- The optional keyword controls the intercept in the regression-based test.
- TRUE
- The default setting includes the intercept term.
- FALSE
- Excludes the intercept term. When FALSE is specified, at least one variable must be specified after either BY or WITH in the META REGRESSION command.
- DISTRIBUTION
- The optional keyword controls the distribution used in the regression-based
tests. T is the default setting, which estimates the statistics based on the
t distribution. When NORMAL is specified, the statistics are estimated
based on the normal distribution. The process is available when
/INFERENCE ADJUSTSE = NONE
, and is otherwise ignored. - ESTIMATE
- The optional keyword specifies the estimator. The process is available when
MODEL = RANDOM
, and is otherwise ignored. When /TRIMFILL is specified, this keyword also controls the estimator that is used by pooling in the trim-and-fill analysis. When /BIAS is specified, this keyword also controls the estimator that is used by the regression-based test.- REML
- The default setting that applies the iterative method and calculates the restricted maximum likelihood estimator.
- ML
- Applies the iterative method and calculate the maximum likelihood estimator.
- BAYES
- Applies the iterative method and calculate the empirical Bayes estimator.
- HEDGES
- Applies the non-iterative method and calculates the Hedges estimator.
- HUNTER_SCHMIDT
- Applies the non-iterative method and calculates the Hunter-Schmidt estimator.
- DERSIMONIAN_LAIRD
- Applies the non-iterative method and calculates the DerSimonian-Laird estimator.
- SIDIK_JONKMAN
- Applies the non-iterative method and calculates the Sidik-Jonkman estimator.
- ADJUSTSE
- The optional keyword controls whether to apply the Knapp-Hartung standard
error adjustment. The process is available when
MODEL = RANDOM
, and is otherwise ignored. When /TRIMFILL is specified, this keyword also controls the standard error adjustment that is used by pooling in the trim-and-fill analysis. When /BIAS is specified, this keyword also controls the standard error adjustment that is used by the regression-based test.- NONE
- The default setting that does not apply the adjustment.
- KNAPP_HARTUNG
- Applies the adjustment method.
- TRUNCATED_KNAPP_HARTUNG
- Applies the adjustment method and truncates the value if less than 1 when estimating the variance-covariance matrix.