A highest posterior density interval for examining shifts in individual regression coefficients between subsamples with different error variances is derived from a weighted-t posterior density. Compared to a double-t approach, the weighted-t approach is computationally more efficient and allows a direct comparison to a sampling theory approach. In a numerical example the weighted-t, double-t and confidence interval approaches are compared.
All Science Journal Classification (ASJC) codes
- Statistics and Probability
- density intervals weighted-t
- double-t densities
- highest posterior