Linear regression using a hierarchical bayesian model with gaussian mixture modeling of the independent variable, as in Kelly (2007)
https://arxiv.org/abs/0705.2774
The method incorporates:
- heteroscedastic and possibly correlated measurement errors in both the dependent and independent variables
- intrinsic scatter in the dependent variable
- non-detections/upper limits
- selection effects on the measurements.