search

tom-hc-park / MSc-RA-Bayesian-evidence-synthesis

Research Project at M.Sc. Statistics at University of British Columbia
4 stars 0 forks source link

the idea of the generative model #10

Open tom-hc-park opened 4 years ago

tom-hc-park commented 4 years ago

generative model pic

This is what I believe the generative model for the project...

P.S. I just realized that I omitted the survey data (x_a ~ bin(n_a, p_a))

I wonder if I need to show how the joint distribution of every variables (LHS) can be shown as the product of certain distributions in appendix section (RHS)?

paulgstf commented 4 years ago

Yes, I think that works.

In general one is always starting with a generic factorization f(a)f(b|a)f(c|a,b)f(d|a,b,c)... but then making conditional independence assumptions (i.e. crossing out some quantities to the right of the conditioning bar) in order to have something scientifically sensible for the problem at hand.