[azalemeth]

I think this article is a great opportunity to mention two under-used statistical techniques: Deming regression [0] and the Theil-Sen estimator [1].

They both fit straight lines to noisy data.

Deming regression is an errors-in-variables model that tries to fit the line of best fit when you have errors in both x and y and they are both known and in general _different_; the Theil-Sen estimator is based on medians and is particularly robust if you have an error process that fails more "one way" than the other. Simple linear regression is everywhere in our lives and yet remarkably not robust to errors that are not IID normal, particularly with a small number of data points: a process that can only fail in one direction if it breaks is likely to completely and utterly bugger up the line that you fit. Both approaches have their place and I wish were more widely used, particularly by people who like fitting linear models to complex phenomena because they are easily understood.

[0] https://en.wikipedia.org/wiki/Deming_regression [1] https://en.wikipedia.org/wiki/Theil%E2%80%93Sen_estimator