MSE FINDR

MSE FINDR estimates mean square error (within-study residual variance) from treatment means and post hoc test results in ANOVA-type experimental designs to enable inclusion of studies lacking explicit variance estimates in meta-analyses.

Key Features:

• Experimental Design Support: Handles Latin square, completely randomized, randomized complete block, two-way factorial, and split-plot designs.
• Post Hoc Test Integration: Uses Fisher's least significant difference (LSD), Tukey's honest significant difference (HSD), Bonferroni, Sidak, and Scheffe tests.
• Input Requirements: Processes treatment means, significance level (alpha), number of replicates, and post hoc mean separation test results provided via CSV file.
• Variance Recovery: Recovers within-study residual variance (MSE) based on user-provided study information for one-way and two-way experimental designs.
• Output Export: Recovered within-study variance can be downloaded and exported as a CSV file.

Scientific Applications:

• Meta-Analysis Inclusion: Enables inclusion of studies that lack explicit variance estimates but report appropriate summary statistics, preventing small-study effects and loss of statistical power in research synthesis.
• Effect Size Estimation: Supports meta-analyses involving estimates of effect sizes based on a continuous response by providing recovered within-study variance.

Methodology:

MSE FINDR recovers mean square error using commonly reported information on treatment means, significance level, number of replicates, and post hoc mean separation tests; simulations validated that recovered MSE accurately predicted actual ANOVA MSE for one-way designs when summary statistics were available for the single factor, and for two-way designs and split-plot designs when summary statistics were available for both factors and the sub-plot factor, irrespective of the post hoc mean separation test.