Reading a Signal in 3D — the Forest Plot

Everything Lesson 7.3 explained about point estimates and confidence intervals mattering together is exactly what a forest plot is built to show at a glance. Rather than reading a table of numbers row by row and mentally tracking which ratios have narrow intervals and which don’t, a forest plot renders each drug-event pair as a single vertical bar: the point estimate marked clearly, the bar spanning the confidence interval’s lower to upper bound, and a reference line marking the null value (ROR = 1, meaning no disproportionate reporting at all).

Reading it is genuinely fast once you know what to look for. A bar sitting entirely above the null line — meaning even its lower confidence bound clears 1 — is a real candidate for further investigation, and the higher and tighter that bar sits, the stronger the candidate. A bar that straddles the null line, even if its point estimate looks impressively high, is telling you the data can’t yet distinguish this pair from background noise; the true value is statistically plausible anywhere within that wide interval, including "no elevated risk at all."

This is also why forest plots are such a natural companion to the scatter-style Signal Cluster visualization introduced in Module 3: the scatter view is excellent for surveying many pairs at once, comparing case volume against ratio magnitude broadly; the forest plot is built specifically to make the point-estimate-versus-uncertainty relationship for a smaller, focused set of candidate pairs immediately visible, which is exactly the comparison a reviewer needs once they’ve narrowed down to a handful of pairs worth a closer look.

Explore the pairs in this lesson’s 3D forest plot directly — click through them and notice how case count relates to interval width, and how a pair’s color reflects whether its interval clears the null line. That relationship — few cases, wide interval, harder to trust; many cases, tight interval, easier to trust — is the single most important intuition disproportionality analysis is built to teach.