A case study in techniques of allocation.

A number of techniques for allocating an individual into one of two groups have been compared in a practical situation. On the basis of the answers to nine Yes/No questions it was necessary to allocate patients previously treated for thyrotoxicosis into those suspected to be hypothyroid and those thought to be euthyroid. Eight different methods have been applied to this problem and the main conclusion from the analysis is that seven of the techniques give broadly similar results, including the simple one of symptom counting. The other method, the multinomial, suffered from the need to estimate a large number of parameters on a relatively small set of data, and the resulting allocation rule performed badly in an independent sample.