Small bug when calling results.posteriors.asini()

[cevawm]

If you try to retrieve asini from the posteriors of a RVHGPMmodel model run, and specifically when you provide the posterior_holder method asini() with a primary star mass and a +-1 sigma value (i.e. if you run res.posteriors.asini(star_mass=[tar_prim_mass, tar_prim_mass_err])), you get an error in the function get_planet_semimajor_axis():

`line 370, in get_planet_semimajor_axis
    a = f * star_mass**(1. / 3) * (P / (2 * np.pi))**(2. / 3)
        ~~~~~~~~~~~~~~~~~~~~~~~~^~~~~~~~~~~~~~~~~~~~~~~~~~~~~
ValueError: operands could not be broadcast together with shapes (167876,) (83938,2) `

This does not happen if you provide just a mass value without a +-1 sigma, i.e. if you only provide a float, not a tuple. In other words, the function has a small bug when dealing with a tuple vs a float for the primary star mass.

I made a quick fix of this error by replacing line 369 in analysis.py with the following:

`star_mass = np.random.normal(star_mass[0], star_mass[1], P.shape)`

(i.e., I replaced line 369 with line 197, which is from the similar get_planet_mass() function).

[j-faria]

Thank you for reporting and suggesting a fix, @cevawm!
It should be fixed now and should appear in 6.3.0. I'll keep this open until I write a test.

[pedrorfigueira]

Hello!

I came across a very similar problem using v6.3.0, when calculating detection limits while providing stellar errors:

    M, A = get_planet_mass_and_semimajor_axis(P, K, E, star_mass=star_mass,
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/pedro/anaconda3/envs/RenV/lib/python3.12/site-packages/kima/pykima/analysis.py", line 446, in get_planet_mass_and_semimajor_axis
    star_mass = np.repeat(star_mass.reshape(-1, 1), P.shape[1], axis=1)
                                                    ~~~~~~~^^^
IndexError: tuple index out of range

My fix was to check dimensions of P to handle 1D vs 2D and apply @cevawm fix. I also removed the newly introduced is instance check at the end otherwise I get types mismatches

    M, A = get_planet_mass_and_semimajor_axis(P, K, E, star_mass=star_mass,
    ^^^^
ValueError: too many values to unpack (expected 2)

The code becomes then:

def get_planet_mass_and_semimajor_axis(P, K, e, star_mass=1.0,
                                       full_output=False, verbose=False):
    """
    (...)
    """

    if verbose:
        print('Using star mass = %s solar mass' % star_mass)

    if isinstance(star_mass, tuple) or isinstance(star_mass, list):
        # include (Gaussian) uncertainty on the stellar mass
        star_mass = star_mass_samples(*star_mass, P.shape[0])
        # ---- FIX: handle 1D vs 2D P safely ----
        if np.ndim(P) == 1:
            star_mass = star_mass.reshape(-1, 1)
        else:
            star_mass = np.repeat(star_mass.reshape(-1, 1), P.shape[1], axis=1)
        # -------------------------------------

    mass = get_planet_mass(P, K, e, star_mass, full_output)
    a = get_planet_semimajor_axis(P, K, star_mass, full_output)

    return mass, a

Let me know what you think.

[j-faria]

I hope it's fixed now 🤞
I made the function always return the star mass; probably not ideal but it's the best solution I can think of.

[pedrorfigueira]

Thanks for the fix!
It worked correctly for the different use cases I tried.