through_origin parameter is not working correctly in DoasCalibData.fit_calib_polynomial()

The parameter through_origin does not force the fit through the origin when few data is available. In the below figure I used DoasCalibData.fit_calib_polynomial with through_origin=True (blue curve) and a simple least square fit (red curve, code below). The blue curve does not go through zero whereby it is different than the non-forced curve (see below). I suggest that you replace the additional zero-data with a simpler fit model or a different function.

With through_origin=True figure_5_lstsq

With through_origin=False figure_5

def fit_linear(calib, plot=False):
    """ Simple linear fit with forced origin """
    tau = calib.tau_vec
    scd = calib.doas_vec
    tau = tau[:,np.newaxis]
    a, res, rank, sgl = np.linalg.lstsq(tau, scd)
    if plot:
        fig, ax = plt.subplots()
        plt.plot(tau, scd, 'bo')
        plt.plot(tau, a*tau, 'r-', label='forced origin')
        ax.set_xlabel(r'$\tau$')
        ax.set_ylabel(r'SO$_2$ SCD [molec/cm$^{2}$]')
        ax.grid()
        plt.show()
    return a

jgliss / pyplis

through_origin parameter is not working correctly in DoasCalibData.fit_calib_polynomial() #3