I've been trying to use the Poisson zeros trick to model a skewed normal distribution. The code runs and returns a result. In order to make sure that it is working correctly, I've used rsn (random skewed normal) to create a data set with known parameters and then I see how closely the model comes to yielding those known parameters. I've used sample sizes of 50 and 500 and modified the priors and modeling parameters multiple times. (I've even "cheated" and set the initial values to the correct values, etc.) But the outputs do not seem close to me.
My parameters are:
Location = 0
Scale = 1
Shape = 0.5
Across multiple modeling runs, here are the parameter values I'm getting:
Location: Average = 0.63 (Std = 0.29)
Scale: Average = 0.48 (Std = 0.08)
Shape: Average = 0.98 (Std = 0.12)
For example, are these values actually the best I should expect?
Of course, it's possible that my equation is wrong, although I've checked it many times (and had someone else check it). So, I doubt this is the problem. But I'm willing to consider anything...
Any/all advice/suggestions would be most welcome! Thanks!
Can you post the code please?
Log in to post a comment.