ideas for straightline from Froeschle

[davidwhogg]

• the scale dependency of the Bayes Factor (i.e. the evidence) when using the line angle instead of the slope as a parameter if the data points have units.
• the issue of hitting the lower limits of the uncertainties in both directions with the posterior probability distribution, because the model does not know where to put the scatter. I showed you a corner plot with this effect. You suggested to give not only upper limits but also sensable lower limits to the uncertainties.
• the question of Markus about how one can put into a model the difference between all data points lying on one side of the line or equally distributed on both sides of the line. In our case we couldn't tell our model that it is an important finding that all redshifts lie above zero.

[davidwhogg]

ps. see email from 2016; and thank him