@mattiasthalen
Thanks for the reply! Yeah as you posted x_0 = F(x_1 -x_2 - dp) + C, so are you suggesting to replace y with x0 in the code?
Please correct me if I am wrong, then wouldn’t replacing y with x0 result in P(C, F, dp , x_0| x_1, x_2) = \frac {P(x_1, x_2 | C, F, dp, x_0) \times P(C, F, dp, x_0)} { P(x_1, x_2)}?
If this is right, then it deviates from the actual form P(C, F, dp | x_0, x_1, x_2) = \frac {P(x_0, x_1, x_2 | C, F, dp) \times P(C, F, dp)} { P(x_0, x_1, x_2)}, right?
My question is about training a model that could the most credible regions for the combination of C,F,dp given each combination of x0,x1,x2, my current code is looking at all the data and then give me a posterior distribution for C,F,dp with observing all the data x0,x1,x2. I could loop through the data (x0,x1,x2) and for each point run the code to find the most probable C,F,dp, after I found posterior distribution for each of C,F,dp given all the data points and then predict one by one what would be the C,F,dp for each. Do you think this is the right way to find the most probable combination for each example of dataset?