Meta-Analysis Binary

The Meta-Analysis Binary procedure performs meta-analysis with binary outcomes on raw data that are provided in the active dataset for the estimation of the effect size.

Refer to the following introductory video for a brief overview on the Meta-Analysis Binary procedure:

Example
Several research studies were conducted in history to investigate a faddish but debatable medicine to help treat type II diabetes. The oral medicine was claimed to be able to reduce the blood glucose level after meals. Data were collected from different research sites from 1979 to 1986.
A principal investigator would like to draw statistical inference about the effect of the oral medicine. Due to the fact that the data were generated from different studies, she proposed the idea of synthesizing the results across the studies to reach an overall understanding of the effect and to identify those underlying sources of variation in the outcomes.
Statistics
Confidence interval, Log Odds Ratio, Peto's Log Odds Ratio, Log Risk Ratio, risk difference, random effects, fixed effects, inverse variance, Mantel-Haenszel, iterations, step-halving, convergence, cumulative statistics, cumulative effect size, restricted maximum likelihood, REML, maximum likelihood, ML, Empirical Bayes, Hedges, Hunter-Schmidt, DerSimonian-Laird, Sidik-Jonkman, Knapp-Hartung, Egger's Test, Harbord's Test, Peters' Test, intercept in regression, dispersion parameter, homogeniety, heterogeniety, exponentiated statistics, standard error, p-value, study weight.

Obtaining a Meta-Analysis Binary analysis

  1. From the menus choose:

    Analyze > Meta Analysis > Binary Outcomes > Raw Data...

  2. Under the Treatment Group section, select a Success variable to represent the “success” counts for the treatment group. The selected variable must be numeric (string variables are not supported).
  3. Select a Failure variable to represent the “failure” counts for the treatment group. The selected variable must be numeric (string variables are not supported).
  4. Under the Control Group section, select a Success variable to represent the “success” counts for the control group. The selected variable must be numeric (string variables are not supported).
  5. Select a Failure variable to represent the “failure” counts for the control group. The selected variable must be numeric (string variables are not supported).
  6. Optionally, select Study ID and/or Study Label variables. The selected Study IDvariable cannot be the same as the selected Study Label variable.
  7. Optionally, select an Effect Size setting. Available options are Log Odds Ratio, Peto's Log Odds Ratio, Log Risk Ratio, and Risk Difference.
  8. Optionally, select a Model setting. When Trim-and-Fill settings are enabled, the setting also controls the model that is used by pooling in the trim-and-fill analysis. When Bias settings are enabled, the setting also controls the model that is used by the regression-based test.
    Random-effects
    The default setting builds the random-effects model.
    Fixed-effects
    Builds the fixed-effects model. Inverse-variance estimates the inverse-variance weight. Mantel-Haenszel estimates the Mantel-Haenszel weight.
  9. Optionally, you can:
    • Click Criteria... to specify the general criteria.
    • Click Analysis to specify the subgroup and cumulative analysis.
    • Click Inference to specify the estimation methods.
    • Click Contrast to control the contrast test.
    • Click Bias to access the publication bias by conducting the EGGER’s regression-based test.
    • Click Trim-and-Fill to implement the trim-and-fill analysis of publication bias.
    • Click Print to control the table outputs.
    • Click Save to save the estimated statistics to the active data set.
    • Click Plot to specify which plots to include in the output.
  10. Click OK.

This procedure pastes META BINARY command syntax.