Local covariance matrix

Hello,

Please have a look at Figure 5 of Czekala et al. 2015:

It explains how the addition of local kernel can be so useful in some cases. My question is: how can I add an outlier line (for example, 5202.2 Angestrum) into code?

As the Starfish documentation says, I have to define its kernel by its hyperparameters:

log_amp: The natural logarithm of the amplitude of the kernel
mu: The location of the local kernel
log_sigma: The natural logarithm of the standard deviation of the kernel

Does it mean that I use the following definition in the SpectrumModel function as:

local_cov = dict(log_amp:?, mu: 5202.2, log_sigma:?)

If yes, please guide me on how to select "log_amp" and "log_sigma". Does "mu" value is correct?

And how to define them in the prior values:

"local_cov:log_amp": st.norm(?),
"local_cov:log_sigma": st.norm(?),
"local_cov:mu": st.norm (5222.1, 0.1)

Thanks a lot for any help in advance.

Start with the Starfish documentation's local-kernel section and the SpectrumModel and prior definitions referenced in the issue. Clarify how an outlier line is represented, how its hyperparameters and priors should be selected, and update the relevant usage guidance with a concrete example.