""" Hard-coded results for test_regression """ # REGRESSION MODEL RESULTS : OLS, GLS, WLS, AR import numpy as np class Longley(object): ''' The results for the Longley dataset were obtained from NIST http://www.itl.nist.gov/div898/strd/general/dataarchive.html Other results were obtained from Stata ''' def __init__(self): self.params = ( 15.0618722713733, -0.358191792925910E-01, -2.02022980381683, -1.03322686717359, -0.511041056535807E-01, 1829.15146461355, -3482258.63459582) self.bse = ( 84.9149257747669, 0.334910077722432E-01, 0.488399681651699, 0.214274163161675, 0.226073200069370, 455.478499142212, 890420.383607373) self.scale = 92936.0061673238 self.rsquared = 0.995479004577296 self.rsquared_adj = 0.99246501 self.df_model = 6 self.df_resid = 9 self.ess = 184172401.944494 self.ssr = 836424.055505915 self.mse_model = 30695400.3240823 self.mse_resid = 92936.0061673238 self.mse_total = (self.ess + self.ssr)/(self.df_model + self.df_resid) self.fvalue = 330.285339234588 self.llf = -109.6174 self.aic = 233.2349 self.bic = 238.643 self.pvalues = np.array([ 0.86314083, 0.31268106, 0.00253509, 0.00094437, 0.8262118, 0.0030368, 0.0035604]) # pvalues from rmodelwrap self.resid = np.array(( 267.34003, -94.01394, 46.28717, -410.11462, 309.71459, -249.31122, -164.04896, -13.18036, 14.30477, 455.39409, -17.26893, -39.05504, -155.54997, -85.67131, 341.93151, -206.75783)) # Obtained from R using # m$residuals / sqrt(sum(m$residuals * m$residuals) / m$df.residual) self.resid_pearson = np.array(( 0.87694426, -0.30838998, 0.15183385, -1.34528175, 1.01594375, -0.81780510, -0.53812289, -0.04323497, 0.04692334, 1.49381010, -0.05664654, -0.12811061, -0.51024404, -0.28102399, 1.12162357, -0.67821900)) def conf_int(self): # a method to be consistent with sm return [ (-177.0291, 207.1524), (-.111581, .0399428), (-3.125065, -.9153928), (-1.517948, -.5485049), (-.5625173, .4603083), (798.7873, 2859.515), (-5496529, -1467987)] HC0_se = (51.22035, 0.02458, 0.38324, 0.14625, 0.15821, 428.38438, 832212) HC1_se = (68.29380, 0.03277, 0.51099, 0.19499, 0.21094, 571.17917, 1109615) HC2_se = (67.49208, 0.03653, 0.55334, 0.20522, 0.22324, 617.59295, 1202370) HC3_se = (91.11939, 0.05562, 0.82213, 0.29879, 0.32491, 922.80784, 1799477) class LongleyGls(object): ''' The following results were obtained from running the test script with R. ''' def __init__(self): self.params = (6.73894832e-02, -4.74273904e-01, 9.48988771e+04) self.bse = (1.07033903e-02, 1.53385472e-01, 1.39447723e+04) self.llf = -121.4294962954981 self.fittedvalues = [ 59651.8255, 60860.1385, 60226.5336, 61467.1268, 63914.0846, 64561.9553, 64935.9028, 64249.1684, 66010.0426, 66834.7630, 67612.9309, 67018.8998, 68918.7758, 69310.1280, 69181.4207, 70598.8734] self.resid = [ 671.174465, 261.861502, -55.533603, -280.126803, -693.084618, -922.955349, 53.097212, -488.168351, 8.957367, 1022.236970, 556.069099, -505.899787, -263.775842, 253.871965, 149.579309, -47.873374] self.scale = 542.443043098**2 self.tvalues = [6.296088, -3.092039, 6.805337] self.pvalues = [2.761673e-05, 8.577197e-03, 1.252284e-05] self.bic = 253.118790021 self.aic = 250.858992591 class CCardWLS(object): def __init__(self): self.params = [ -2.6941851611, 158.426977524, -7.24928987289, 60.4487736936, -114.10886935] self.bse = [ 3.807306306, 76.39115431, 9.724337321, 58.55088753, 139.6874965] # NOTE: we compute the scale differently than they do for analytic # weights self.scale = 189.0025755829012 ** 2 self.rsquared = .2549143871187359 self.rsquared_adj = .2104316639616448 self.df_model = 4 self.df_resid = 67 self.ess = 818838.8079468152 self.ssr = 2393372.229657007 self.mse_model = 818838.8079468152 / 4 self.mse_resid = 2393372.229657007 / 67 self.mse_total = (self.ess + self.ssr) / 71. self.fvalue = 5.730638077585917 self.llf = -476.9792946562806 self.aic = 963.95858931256 self.bic = 975.34191990764 # pvalues from R self.pvalues = [ 0.4816259843354, 0.0419360764848, 0.4585895209814, 0.3055904431658, 0.4168883565685] self.resid = [ -286.964904785, -128.071563721, -405.860900879, -20.1363945007, -169.824432373, -82.6842575073, -283.314300537, -52.1719360352, 433.822174072, -190.607543945, -118.839683533, -133.97076416, -85.5728149414, 66.8180847168, -107.571769714, -149.883285522, -140.972610474, 75.9255981445, -135.979736328, -415.701263428, 130.080032349, 25.2313785553, 1042.14013672, -75.6622238159, 177.336639404, 315.870544434, -8.72801017761, 240.823760986, 54.6106033325, 65.6312484741, -40.9218444824, 24.6115856171, -131.971786499, 36.1587944031, 92.5052108765, -136.837036133, 242.73274231, -65.0315093994, 20.1536407471, -15.8874826431, 27.3513431549, -173.861785889, -113.121154785, -37.1303443909, 1510.31530762, 582.916931152, -17.8628063202, -132.77381897, -108.896934509, 12.4665794373, -122.014572144, -158.986968994, -175.798873901, 405.886505127, 99.3692703247, 85.3450698853, -179.15007019, -34.1245117188, -33.4909172058, -20.7287139893, -116.217689514, 53.8837738037, -52.1533050537, -100.632293701, 34.9342498779, -96.6685943604, -367.32925415, -40.1300048828, -72.8692245483, -60.8728256226, -35.9937324524, -222.944747925] self.resid_pearson = [ -0.90569581, -0.75496938, -1.28663890, -0.11309411, -0.24746253, -0.47181831, -1.02062293, -0.31403683, 1.62862142, -0.84973225, -0.42919669, -0.78007426, -0.63913772, 0.29787637, -0.38364568, -0.21381846, -0.85577361, 0.54156452, -0.48496031, -1.15374603, 0.41145856, 0.23996158, 2.70305838, -0.53171027, 0.79057028, 1.82433320, -0.04150362, 0.97048328, 0.13667658, 0.26750667, -0.12690810, 0.11703354, -0.72689772, 0.34160874, 0.71332338, -0.75079661, 1.73137185, -0.39477348, 0.04107215, -0.11332274, 0.22952063, -0.88580496, -0.67239515, -0.17656300, 4.48867723, 2.61499898, -0.16988320, -0.63136893, -0.68135396, 0.06351572, -0.64467367, -0.37800911, -0.64304809, 1.88607184, 0.57624742, 0.60875207, -0.78636761, -0.17897383, -0.21716827, -0.07885570, -0.57566752, 0.25202879, -0.29176531, -0.54378274, 0.30203654, -0.57460072, -0.72378394, -0.23853382, -0.17325464, -0.24121979, -0.10269489, -0.57826451] def conf_int(self): # a method to be consistent with sm return [ (-10.2936, 4.90523), (5.949595, 310.9044), (-26.65915, 12.16057), (-56.41929, 177.3168), (-392.9263, 164.7085)] class LongleyRTO(object): def __init__(self): # Regression Through the Origin model # from Stata, make sure you force double to replicate self.params = [ -52.993523, .07107319, -.42346599, -.57256869, -.41420348, 48.417859] self.bse = [ 129.5447812, .0301663805, .4177363573, .2789908665, .3212848136, 17.68947719] self.scale = 475.1655079819532**2 self.rsquared = .9999670130705958 self.rsquared_adj = .9999472209129532 self.df_model = 6 self.df_resid = 10 self.ess = 68443718827.40025 self.ssr = 2257822.599757476 self.mse_model = 68443718827.40025 / 6 self.mse_resid = 2257822.599757476 / 10 self.mse_total = (self.ess + self.ssr) / 16. self.fvalue = 50523.39573737409 self.llf = -117.5615983965251 self.aic = 247.123196793 self.bic = 251.758729126 self.pvalues = [ 0.6911082828354, 0.0402241925699, 0.3346175334102, 0.0672506018552, 0.2263470345100, 0.0209367642585] self.resid = [ 279.902740479, -130.324661255, 90.7322845459, -401.312530518, -440.467681885, -543.54510498, 201.321121216, 215.908889771, 73.0936813354, 913.216918945, 424.824859619, -8.56475830078, -361.329742432, 27.3456058502, 151.28956604, -492.499359131] # Obtained from R using # m$residuals / sqrt(sum(m$residuals * m$residuals) / m$df.residual) self.resid_pearson = [ 0.58906369, -0.27427213, 0.19094881, -0.84457419, -0.92697740, -1.14390695, 0.42368630, 0.45438671, 0.15382784, 1.92189233, 0.89405658, -0.01802479, -0.76042924, 0.05754964, 0.31839340, -1.03647964] def conf_int(self): return [ (-341.6373, 235.6502), (.0038583, .1382881), (-1.354241, .5073086), (-1.194199, .0490617), (-1.130071, .3016637), (9.003248, 87.83247)]