Interactive forest plots: the effect of introducing alternative proteins on meat purchases, across 7 restaurants in Tier 1 and 19 restaurants in Tier 2 (the 19 include the 7).
Tier 1 and Tier 2, analyses A1 to A6. Restaurants ordered by effect size.
The same estimates, with per-restaurant colours, inline names and a numbered legend.
Estimates are reported as relative rate ratios (RRR), that is, the effect of the exposure on the outcome pooled across restaurants divided by the effect of the exposure on total purchases pooled across restaurants. Estimates are pooled across restaurants, and so represent the average effect across restaurants in the set; the accompanying credible intervals (CI) reflect uncertainty about the average across restaurants generally, inside and outside the set. The RRRs are estimated by Monte Carlo division over the posterior distributions of the effect estimates from two separate models, the outcome model and the total purchases model, computed draw by draw, so the point estimates and credible intervals reported here are summaries (namely, the median and quantiles) of the resulting RRR distribution.
If the outcome were meat purchases, for example, RRR = 1.05 would indicate a 5% higher exposure-driven growth rate in meat purchases than in total purchases, while RRR = 0.95 would indicate a 5% lower rate. If total purchases doubled (200% of baseline) and meat purchases grew 1.9-fold (190% of baseline), the exposure's RRR estimate would be 190/200 = 0.95, indicating a relative decline in meat purchases. In cases when total purchases are falling, an RRR above 1 indicates a slower decline in meat purchases and an RRR below 1 indicates a faster decline.