refine bail policy vs crime rate results

[akelleh]

If the results of the simpler ticket work out, AND we want more robust results, we can do the following:

To make the analysis more robust, we probably want to take into account both the x and y errors. To do that, you need "total least squares" or "orthogonal least squares" regression. There's an implementation in scipy here, but the API is pretty low-level, which means it's harder to work with.
https://docs.scipy.org/doc/scipy/reference/odr.html

[mayank5695]

I am wondering that there are other methods to account for both x and y errors. We can also get that from scikit-learn library. We can easily use different regression models from there and also account for error. Do let me know what we should do about it?