Discussion about this post

User's avatar
Indy Neogy's avatar

Mildly fascinated since I'm mixed race and from Yorkshire (although family full of medics rather than dentists) just how many mixed race dentists there are in Yorkshire. My gut instinct is not many and also not easy to find out.

Philip Koop's avatar

Tangential to the spirit of your post, "Type 3 Error" reminds me of Gelman & Carlin's "Type M Error" (conditioned on being "statistically significant, a result exaggerates the effect size) and even worse "Type S Error" (conditioned on being statistically significant, a result has the opposite sign of the true effect).

These *sound* like social errors, and ultimately they are (because it can be expensive to run a study with sufficient power, and because publication is biased toward sensational results.) But proximately, they are just the impersonal mathematical selection effect of the significance test on underpowered studies.

8 more comments...

No posts

Ready for more?