specifying prior precision parameter g in bayesian regression

[Mahmoud Arayssi]

I am trying to specify prior values for constant and slope parameters in Bayesian regression. If I choose igamma distribution with shape =1 and scale 0.5 , I specify -1.32 for initial constant parameter and 0 for initial slope parameter; in order to generate posterior distribution, how can I specify g (prior precision parameter) ? There was no option to do this . I am trying to fit a logit transformation model of consumption to savings ratio on Log of disposable income. The slope parameter should be negative to show a departure from the proportionality of consumption to income (Friedman hypothesis)

[Red Owl]

Mahmoud,

I believe you may want the Zellner's g-prior option.

To read about this, type

help bayes

Then open the bayes.pdf manual from the top of the online help screen for bayes and search for Zellner. Zellner's g-prior is discussed in several places in the bayes.pdf manual.

I hope that's what you need.

Red Owl
Stata 15

[Mahmoud Arayssi]

Thanks that was helpful. Too bad zellnerg is not in the dropdown menu...

[Mahmoud Arayssi]

How do I check if Bayesian regression is converging after iterations?