Hi everyone again,
I am writing because I am trying to figure out how to handle point 3 of my previous reply to @AlexAndorra suggestion:
I would like to write down a proper Likelihood taking into account that my data is truncated (to be honest, I think that “truncation” is not the correct word here because all the obs < 0 are set zero).
I am modelling the observations with a Normal Likelihood, but, as Alex made me notice, my observations can’t be negative.
I’ve checked the solution for censored data and for truncated data but I think that my situation is different, given that all my negative observations are not truncated, but they are set to zero so that the histogram of my observation shows a high obs=0 bin.
How can I make the model guess which obs=0 is really 0 and which one is more likely to be negative, imputed to 0? Obviously I’m referring to the model that is in my first post here.
Thank you!