All stars have radius of 0.98

[elisabethadams]

(Well, they probably don't actually.)

But the following instructions result in essentially the same values for Teff from 3000-7000:

mist = MIST_Isochrone()
model  = StarModel(mist, Teff=(5770, 80), logg=(4.44, 0.08), feh=(0.,0.1))
model.fit()
model.samples['radius_0_0'].quantile([0.15,0.5,0.85])

0.15    0.922251
0.50    0.982562
0.85    1.064326

[timothydmorton]

The secret of the universe has been discovered!

Two possible thoughts:

• Is the model.fit() call actually running a new fit every time, or is it reading the results of an old fit?
• Are you changing Teff but leaving logg = (4.44, 0.08)? If so, then the logg constraint might dominate over the Teff?

[elisabethadams]

Definitely running a new model.fit() every time (the numbers change after the 3rd decimal place every time I run it).

Changing logg (and increasing the errors) doesn't really affect it:

model  = StarModel(mist, Teff=(3000,200), logg=(4.7, 0.28), feh=(0.,0.1))

0.15    0.921389
0.50    0.981665
0.85    1.061877

[timothydmorton]

That's definitely not right...do the new fits take time to run? What if you call model.fit(refit=True)? You're using v1.0, right?

[timothydmorton]

(in other words I wonder whether the differences are just from choosing different random samples from the already-run multinest chains)

[elisabethadams]

Hmm, the refit=True may have done the trick! Why is refit=False the default?

[timothydmorton]

In my use, it's much more common for me to want to quickly pull up the results of an already-fit model than to change the parameters of a model and then refit. You can avoid this kind of problem by giving your StarModels names (use the name= keyword), because that will define what the multinest chains are called, and thus you won't have clashing fit files.

There is definitely an area that needs improvement! Thanks for bringing it to my attention.