www.cmm.bristol.ac.uk/forum

Hi,

While I was looking for information about the p-value for MCMC estimation, I came across the following link:

https://www.cmm.bristol.ac.uk/forum/vie ... 57d6#p3459

Does it mean that we have to use the one-tailed p-value to examine the significance of predictor variables when we use the MCMC approach? I have read through the online course materials, but I didn't see anything about halving the p-value if I use the MCMC approach in MLwiN. I am quite confused about that.

Thanks a lot.

Hi Vivian,

While the one-sided p-values are displayed by default, if you instead specify the zratio option you will get "conventional" p-values based on assuming the parameters have normal sampling distributions. These p-values are associated with two-sided hypothesis testing.

Best wishes

George