9. Consider data on Leptograpsus crabs dataset as used by Ripley (1996) – see Exercise 3.9.odc. The objective is to classify the sex of the crabs from P = 5 scalar anatomical observations. The training set contains N1 = 80 examples (40 of each sex) and the test set includes N2 = 120 examples. Using MVN discrimination gives correct classification rates (training and test samples respectively) of 93.4 and 94.3%. Adapt the following code

(where G = 1 for males, and 2 for females) to assess the benefit of multivariate Student t classification with differing degrees of freedom between subpopulations (McLachlan and Peel, 1998).

Transcribed Image Text:

model # N1 training cases, N2=N-N1 test cases in C populations for (i in 1:N) {G[i] ~ dcat (pi [1:C]); yli, 1:P] ~ (mu [G[i], ], Pr [G[i],,])} v1 [1: P, 1:P]