Consider \vec{v_x}=m\vec{X}+b and \vec{v_y}=m\vec{Y}+b. Then let r=\mathcal{N}(r_{mean},\sigma), where r_{mean}=v_x/v_y. With little algebra, we can see that r_{mean} = \frac{\vec{X}}{\vec{Y}+b/m} + \frac{1}{m/b\ \vec{Y} +1}. \tag{1} So we see that we can infer b/m. The stan code is data { int…

What does it mean if inference is only sometimes working?

FJCC April 25, 2020, 10:10pm 6

I saw the fit failing to converge 1 out of 5 times with N = 20 and most of the time with N = 200, so I do not think the failures you saw are a fluke.

Testing for Identifiability

Topic		Replies	Views
Issues with heterogeneously normalized data Modeling techniques , fitting-issues , specification	6	549	April 23, 2020
Bayesian Inference on Nongenerative Model Modeling techniques , fitting-issues , specification , performance	1	376	February 24, 2023
What Does Stan Calculate in case of Inconsistent Modeling Assumptions Modeling	7	653	September 24, 2019
Model gets stuck - but not quite always Modeling	7	1355	July 18, 2018
Mismatch when predicting inside and outside of STAN Modeling posterior-predictive	4	806	May 20, 2020