search

Starfish-develop / Starfish

Tools for Flexible Spectroscopic Inference
https://starfish.readthedocs.io/en/latest/
BSD 3-Clause "New" or "Revised" License
68 stars 22 forks source link

Best model vs initial model in the single-order spectrum example in v 0.3.0 #129

Closed bersavosh closed 4 years ago

bersavosh commented 4 years ago

Hi there, in the single-order example for v0.3.0, the initial model (prior to MCMC sampling) appears to better fit the data compared to the best fit presumably plotted in the last figure.

I was wondering, is this simply a MAP vs mean issue (given the best fit here is based on mean values)? or caused by something else?

thanks

mileslucas commented 4 years ago

Hi!

That example is currently a work-in-progress. I have coded out what should be an appropriate usage of Starfish but I haven't fine-tuned it to work properly yet. I would suggest using it as a stepping stone for your own MLE process. This example is already a little light so that I can run it in a reasonable time on my laptop.

bersavosh commented 4 years ago

Thanks for the quick response.