Meta-Analysis: Exam 2

Explain how we can estimate systematic variability (heterogeneity).

We can estimate heterogeneity by calculating the total variance from study-to-study, then calculate sampling variance, then we look at the remaining variation and if our total variance is much larger than our sampling variance, then we can assume that there is heterogeneity. You would then estimate between variance.

Explain what the caveats of statistical testing for heterogeneity tests are.

Sometimes a lack of significance can be due to low power/a small number of studies. It can also be due to large within-study variance. Even substantial between-studies dispersion can show non-significance. Also, having a small number of effect sizes can show non-significant results because of low power.

Define I².

It is used to see if the proportion of the observed variance reflects real differences in effect size.

Provide an example of a categorical variable and an example of a continuous variable as moderators.

Explain the challenge of multiple comparisons under meta-regression and provide an alternative option.

Define publication bias.

Define funnel plot, including 1) when we can conclude that there is no publication bias, 2) an advantage, and 3) four disadvantages.

Define trim and fill, including 1) when we can conclude that there is no publication bias, 2) at least one advantage, and 3) two disadvantages.

Define Egger’s regression in RVE, including 1) when we can conclude that there is no publication bias, 2) three advantages, and 3) a disadvantage.

Lab 5: Pooling Effect Size and Test of Heterogeneity

Lab 6: Moderator Analysis

Lab 7: Publication Bias