Good day sir,
Please I need assistance on this !.
I am trying to find priors for some of parameters in my equation. I could not find any from existing studies. I decided to estimate the equations that contain those estimate. After the estimation of regression. I found some of the coefficient to be negative number. my question is is my approach to get prior values correct and can the negative values be use as prior. If yes, which type of probability distribution will I choose for such priors with negative values. thank you.
Your prior must not depend on the data, so your approach looks suspicious. Also, OLS is usually not suitable to get structural estimates. Finally, without knowing more about the parameters it is impossible to give any advice
thank you sir,
Sir, I have some parameters in my models that I did not get prior values for from previous studies. Initially I used OLS to get their values but thank you for the insight provided that OLS is not suitable. Please which method can I use to generate structural estimates for those parameters.
Thank you sir.
That is why you are estimating the model. The prior is your subjective opinion about the parameter before seeing the data. If unsure, use a uniform prior.