Conclusions from brain mapping: need for statistical rigor.

Let us take two groups of, say, 310 people each, drawn from the same population. Now let usdraw two large circles. Withineachcircle, let usdraw31 equallyspacedsmallercir-cles, sothat,foreachsmallcircleinonelargecircle, thereisan equivalently situated small circle in the other large circle. Now we instruct the 31 0 people to form self-assorted sub-groupsof lopeopleeach, andoccupyoneofthesmallcircles. We measure everyone's height, weight, neckcircumference, and abdominal girth. Now we take the mean height, weight, neckcircumference and abdominal girth at each small circle. The next step comprises a comparison of each small circle within one large circle, with the equivalent small circle in the otherlargecircle.Thus,we have31 x4= 124comparisons. If wesetourpvalueforsignificantdifferenceatO.05, we will find, onaverage, 124x0.05= Gsignificantdifferences. Forexam-ple, we might find that circle D1 has a smaller height and weight, circles A1 and F1 have a larger neckcircumference, circle K1 has a larger abdominal girth, and circle 0 1 has a smaller abdominal girth. What are we to make of this data? Wecouldconclude thatthere isadifferencein the mapping of, say, theneckcircumference, withcircle 1 showing largercir-cumference at A1 and F1. Yet, a test for overall statistically significant difference would probably find that there are no real differences here. Going further, we could draw topographical maps of the abdominal girth, assigningdifferentcolors to differentvalues.