Random effects meta regression model (standard error adjustment)

The standard method in the previous example does not take into account the fact that the amount of residual heterogeneity has to be estimated. This can lead to incorrect conclusions about the inference of the significance of the moderator. As a result, the analyst wants to use the Knapp-Hartung method with truncation to adjust the standard error of coefficients. The analyst selects the TRUNCATED_KNAPP_HARTUNG method in the ADJUSTSE subcommand.

META REGRESSION LnRiskRatio with length
/DATA VAR=VARLnRiskRatio
/CRITERIA CILEVEL=95 CLASSMISSING=EXCLUDE MAXITER=200 MAXSTEP=100 CONVERGENCE=0.000001
/INFERENCE MODEL=RANDOM INTERCEPT=INCLUDE DISTRIBUTION=NORMAL ESTIMATE=ML ADJUSTSE=TRUNCATED_KNAPP_HARTUNG
/PRINT COEFF_TEST PARAMETER.

The following section details how to draw statistical inference about the effect, with standard error adjustment, through the Meta-Analysis: Regression procedure.

Re-running the analysis

  1. Recall the Meta-Analysis: Regression analysis dialog:

    Analyze > Meta Analysis > Meta Regression

  2. Click Inference.
    Figure 1. Inference dialog
    Meta-Analysis: Regression Inference dialog
  3. Select the Apply the truncated Knapp-Hartung adjustment settings.

    The Apply the truncated Knapp-Hartung adjustment setting adjusts the standard error of coefficients.

  4. Click Continue.
  5. Click OK.

Model Coefficient Test table

Figure 2. Model Coefficient Test table
Model Coefficient Test table

The model coefficient test is now based on the F test. The significance value is a bit larger than the Wald Chi-square test from the previous example.

Parameter Estimates table

Figure 3. Parameter Estimates table
Parameter Estimates table

The standard errors of coefficient in the Parameter Estimates are larger than those from the user case 1. Thisis expected because the Knapp-Hartung method with truncation adjustment take the uncertainty in between-study estimate into consideration. In this situation, t test is used to compute the sig value instead of z test.

Next