Further improvements to the model

[bdelwood]

A good point was brought up during the presentation today: we could consider correlation between points, motivated by the fact that our data should be correlated through the scanning of the spectrometer (eg systematics in, say, laser power over certain timescales would correlate adjacent data points as the spectrometer scans across channels).

[vnmanoharan]

Consider using MvNormal distribution for the likelihood instead of Cauchy or normal. This would allow you to model the covariance between data points.

[bdelwood]

28 attempts to add a MvNormal likelihood, but it currently doesn't work on our dataset. I left my work in for future use or improvement.