def huber(x): res = sm.RLM(x, np.ones(len(x)), M=norms.HuberT()).fit(scale_est=scale.HuberScale()) return res.params[0]
# m1_Bisquare = RLM(data.endog, data.exog, M=norms.TukeyBiweight()) # results_Bisquare1 = m1_Bisquare.fit() # m2_Bisquare = RLM(data.endog, data.exog, M=norms.TukeyBiweight()) # results_Bisquare2 = m2_Bisquare.fit(cov="H2") # m3_Bisquare = RLM(data.endog, data.exog, M=norms.TukeyBiweight()) # results_Bisquare3 = m3_Bisquare.fit(cov="H3") ############################################## # Huber's Proposal 2 scaling # ############################################## ################ ### Huber'sT ### ################ m1_Huber_H = RLM(data.endog, data.exog, M=norms.HuberT()) results_Huber1_H = m1_Huber_H.fit(scale_est=scale.HuberScale()) # m2_Huber_H # m3_Huber_H # m4 = RLM(data.endog, data.exog, M=norms.HuberT()) # results4 = m1.fit(scale_est="Huber") # m5 = RLM(data.endog, data.exog, M=norms.Hampel()) # results5 = m2.fit(scale_est="Huber") # m6 = RLM(data.endog, data.exog, M=norms.TukeyBiweight()) # results6 = m3.fit(scale_est="Huber") # print """Least squares fit #%s #Huber Params, t = 2. #%s #Ramsay's E Params #%s