What is a meta-analysis?

A meta-analysis pools the results of many studies asking the same question and combines them into one number. Any individual study can land far from the truth on luck alone, so ten of them together say much more than the loudest one does. When a headline claims research now shows something, this is usually the strongest form that claim can take.

Ten trials test the same drug: four find it helps, three find nothing, three are unclear, and every side of the argument now has a study to wave. Counting votes is the tempting move and the wrong one, because it lets a trial of 40 patients outrank a trial of 4,000. A meta-analysis merges them instead, weighting each by how precise it is, so the large careful trials pull hardest on the final answer. The payoff is resolution, since an effect too small for any one trial to confirm becomes visible once thousands of patients are pooled. The limits are real too. Combine studies that measured different things and the answer is about nothing in particular, a problem researchers call mixing apples and oranges. Worse, journals prefer exciting findings, so trials that found nothing often go unpublished, which tilts the visible evidence toward success and has a name, publication bias. A good meta-analysis states its inclusion rules before it starts and tests whether the missing studies could overturn the result.

One neighbour swears the new bakery is excellent and another says it was a letdown, and neither settles anything, because a single visit only catches whatever the kitchen did that afternoon. Read forty reviews and a pattern emerges, especially if you weight the careful regulars above the one line rants. That is the whole idea, not a louder opinion but many weak signals stacked until the shape underneath stops being deniable. The catch is familiar from shopping online, because a bakery that quietly buries its worst reviews leaves you averaging a flattering fiction.