data_explorationsynthesising the literature

Bayesian random-effects meta-analysis

Cultivation effects are never proven by one study, they're aggregates across dozens. A random-effects meta-analysis treats each study's effect as drawn from a distribution centred on the overall effect μ, with between-study heterogeneity τ. This page fits that model on the log-odds scale using a Normal approximation, integrates the posterior numerically, and draws a forest plot.

edit_noteyour studies

Enter each study as successes, total (e.g. for a cultivation indicator: number of heavy viewers overestimating crime out of all heavy viewers studied). One row per line.

bar_chartforest plot

Study effects and the pooled posterior

Studies (k)
Total respondents
Pooled log-odds μ
Pooled proportion
Heterogeneity τ (SD)
95% μ interval

Read the forest plot. Each box is a study: centre = observed proportion, horizontal line = its 95% interval on the log-odds scale. The bottom diamond is the pooled posterior, its vertical width is the uncertainty about the overall effect. The gray shading reflects the between-study heterogeneity τ contributing to each study's effective interval.

Heterogeneity is information. A large τ means effects differ meaningfully across studies, aggregate responsibly, and prefer reporting the prediction interval for a future study over a single pooled number. This is the recommendation consensus in applied meta-analysis (Borenstein et al., 2010; Higgins & Thompson, 2002).