Publication bias: IN CAKE FORM. DataLab at the Curiosity Carnival

This article has been reblogged from the EBM DataLab. See the original post here

We were at Oxford University’s Curiosity Carnival, an event featuring researchers from all disciplines enthusiastically explaining their work to members of the public in an enormous variety of forms, all around the city.

We entered the Great Research Bake-Off and took on the challenge of representing some of the key issues around research integrity through the medium of cake.

We displayed an array of fairy cakes, each representing clinical trials. To produce a good cake, or trial, a good methodology is fundamental. And, of course, the results should be published in a paper (the icing on top); the outcome pre-specified (the colour of the case), and the same outcome reported (same colour flag). A plus or minus on top indicated a positive or negative result. However, only a small number of our cakes looked this good!

The majority were bad trials: cakes which looked odd (poor methodology), had no icing (unpublished), were in a plain case (no pre-specified primary outcome / not pre-registered), had non-matching case and flag (outcome switching), or were very small (small sample size).

The competition was tough – there were many impressive and intricate baked works of art, with prize-winners demonstrating superconductors, primate conservation and autophagy. The cakes were then enjoyed by the carnival-goers and disappeared remarkably quickly!

We hope to be back at the Curiosity Carnival next year, maybe presenting in a different way… if there are any similar events you think we should be at, let us know!