I am taking inverse chi square as my prior distribution for variance of normal sampling distribution. I am little confuse about the choice of value for V0 the parameter of inverse chi prior, that is df and it usually n-1 for one parameter case. If i have n=10 and i choose V0 as 9 it is a reasonable choice for this hyper parameter?


No. That would make the data and the prior equally strong (10 observations strong).

I would recommend using inverse gamma instead.

  • $\begingroup$ Thanks for your suggestion.... But than i have to choose the possible hyper parameter values for IG... Can you recommend me how i can choose these values, if i don't use prior predictive distribution. $\endgroup$ – Tahir Malik Jan 4 '15 at 11:26
  • $\begingroup$ You should use IG(a,b) - match mean and variance to what you want to figure out a and b, see en.wikipedia.org/wiki/Inverse-gamma_distribution $\endgroup$ – TA72 Jan 4 '15 at 17:32

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.