jason-neal / companion_simulations Goto Github PK

Some ideas:

• If companion temp is too low, display warning and default to the 5 lowest spectral models or something.
• Not able to request a spectra that is not available,/ be able to handle that better, not just fail, maybe warn though.

Rather than trying to convert between different names.

Simplify parameter choices using config file, but have some simple defaults.

flag to chose a companion?

or automatically choose largest?

Try F = F * cont model / cont obs from chi**2 parameter paper renormalization of observation

For each observation.

• Inject companion at the correct flux ratio onto the observed spectra
• Try and recover the companion
• Get recovered temperature
• Plot injected temp verse recovered temperature

Nuno suggests making a parabola line his lithium 6 paper.

For each parameter. Pick the chi2 value that is smallest and we should get a parabola.

He also suggest adding a variable normalization that is thrown away at the minimum chi2.
I could add another broadcasted parameter and multiple 1 model or obs by 0.98 and 1.02 and then reduce back to minimum chi 2 for each.

Combine the analysis from the 3 branches into one logical flow. Having a mode/model parameter to choose different code paths.

Need to switch rv and gamma in iam_fake_full_stack.py.

Can I interpolate Multi dimensional data along one axis?

To interpolate to the points of the observation.

Also for the doppler shift of the companion, and system.

DUsty, cond, btsettl etc

The available ones are downloaded onto exo

Try perform the chi square analysis in vectorized form.

Can probably do 3 or 4 dimensions at once.

Find out how to co-add chi2 values. from the same obervation together.

Do this for the enitire gird and then use that chi2 value as the minimum value.

Then average or co-add these across the observations.

broadcast the alpha parameter for the combined modeling.

something like
xy = x + y[:, np.newaxis] * alpha

will result in shape(len(x), len(alpha))

Be able to obtain the spectra with the 4 input parameters. teff feh logg alpha=0.
Be able to iterate over them.

Want to do the models on the outer loops, want to open each one as little as possible.

remove the performance tests and the tests for the old code.

To check that they recover sensible temperatures.

Try use lmfit to match observation to model.

Possibly all parameters. 2 templates, 2 rvs, and an alpha value.
Or else just 1 rv and correct to center of host.

I need to fix the contour plots to let each point be variable for all other parameters.
Not just fixed at the minimum chi squared value.

This will change the confidence levels and the shape hopefully.