Hi all, I have two spectra with wavelength, flux, and error on flux. I want to find out the variability of these two spectra based on the 2 sample Chi-square test. I am using following code:
def compute_chi2_var(file1,file2,zemi,vmin,vmax): w1,f1,e1,c1,vel1 = get_spec_vel(dir_data+file1,zemi) id1 = np.where(np.logical_and(vel1 >= vmin, vel1 < vmax))[0] w2,f2,e2,c2,vel2 = get_spec_vel(dir_data+file2,zemi) id2 = np.where(np.logical_and(vel2 >= vmin, vel2 < vmax))[0] f_int = interp1d(w1[id1], f1[id1]/c1[id1], kind='cubic') e_int = interp1d(w1[id1], e1[id1]/c1[id1], kind='cubic') f_obs,e_obs = f_int(w2[id2]), e_int(w2[id2]) f_exp, e_exp = f2[id2]/c2[id2], e2[id2]/c2[id2] e_net = e_obs**2 + e_exp**2 chi_square = np.sum( (f_obs**2 - f_exp**2)/e_net ) dof = len(f_obs) - 1 pval = 1 - stats.chi2.cdf( chi_square, dof) print('%.10E' % pval) NN = 320 compute_chi2_var(file7[NN],file14[NN],zemi[NN],vmin[NN],vmax[NN]) I am running this code on many files, and I want to grab those pair of spectra where, the p-value of chi-squa is less than 10^(-8), for the change to be unlikely due to a random occurrence. Is my code right concept-wise? Because the chi-squ value is coming out to be very large (positive and negative), such that my p-value is always between 1 and 0 which I know from other's results not correct. Can anyone suggest me is the concept of 2-sample chi-squ applied by me is correct or not? I thank you all in advance. -- https://mail.python.org/mailman/listinfo/python-list