STAT 516

Lab 4

Report due Dec. 12 in my mailbox (earlier is fine)

The following data record the number of still born littermates in 1328 mouse litters. It is likely that these data consist of pooling of several different experiments, so a mixture of binomial models may be called for. Your task is to fit such a model using a Markov chain Monte Carlo approach.It is often found convenient in such analyses to introduce an auxiliary variable Znrl which takes on the value i if the lth litter of size n with r deaths is assigned to mixture component i. Based on analyses of other similar data sets it is reasonable to assume that the model should be a mixture of three components.

n
 r=0
 r=1
 r=2
 r=3
 r=4
 r=5
 r=6
 r=7
 r=8
 r=9
 r=10
1
 15
 1








2
 6
 1
 2







3
 6
 6








4
 7
 2
 3
2





5
 16
 9
 3
 3
 1





6
 57
 38
 17
 2
 2





7
 119
 81
 45
 6
 1

1


8
 173
 118
 57
 16
 3


1

9
 136
 103
 50
 13
 6
 1
 1



10
 54
 51
 32
 5
 1




1
11
 13
 15
 12
 3
 1





12
4
 3
 1






13

1



1