11.07.2015 Views

statisticalrethinkin..

statisticalrethinkin..

statisticalrethinkin..

SHOW MORE
SHOW LESS
  • No tags were found...

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

8.4. CARE AND FEEDING OF YOUR MARKOV CHAIN 263FIGURE 8.9. A chain with unidentifiable parameters, a1 and a2.data=list(y=y) , start=list(a1=0,a2=0,sigma=1) ,chains=2 , iter=1e4 )precis(m8.4)Mean StdDev lower 0.95 upper 0.95a1 -9209.45 3086.31 -13604.71 -4819.27a2 9209.46 3086.31 4819.40 13604.79sigma 1.02 0.05 0.92 1.11ose estimates look suspicious. e means for a1 and a2 are almost exactly the same distancefrom zero, on opposite sides. And the standard deviations of the chains are massive.is is of course a result of the fact that we cannot simultaneously estimate a1 and a2, butonly their sum.Looking at the trace plot reveals more. FIGURE 8.9 shows two Markov chains from themodel above. ese chains do not look like they are stationary, nor do they seem to bemixing very well. Indeed, when you see a pattern like this, it is reason to worry. Don’t usethese samples.Again, weak priors can rescue us. Now the model fitting code is:

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!