# -*- coding: utf-8 -*- """Tests for Regression Diagnostics and Specification Tests Created on Thu Feb 09 13:19:47 2012 Author: Josef Perktold License: BSD-3 currently all tests are against R """ import os import numpy as np import pandas as pd from numpy.testing import (assert_, assert_almost_equal, assert_equal, assert_allclose, assert_array_equal) import pytest from statsmodels.regression.linear_model import OLS from statsmodels.tools.tools import add_constant from statsmodels.datasets import macrodata import statsmodels.stats.sandwich_covariance as sw import statsmodels.stats.diagnostic as smsdia import json import statsmodels.stats.outliers_influence as oi cur_dir = os.path.abspath(os.path.dirname(__file__)) def compare_t_est(sp, sp_dict, decimal=(14, 14)): assert_allclose(sp[0], sp_dict['statistic'], atol=10 ** -decimal[0], rtol=10 ** -decimal[0]) assert_allclose(sp[1], sp_dict['pvalue'], atol=10 ** -decimal[1], rtol=10 ** -decimal[0]) def notyet_atst(): d = macrodata.load(as_pandas=False).data realinv = d['realinv'] realgdp = d['realgdp'] realint = d['realint'] endog = realinv exog = add_constant(np.c_[realgdp, realint]) res_ols1 = OLS(endog, exog).fit() #growth rates gs_l_realinv = 400 * np.diff(np.log(d['realinv'])) gs_l_realgdp = 400 * np.diff(np.log(d['realgdp'])) lint = d['realint'][:-1] tbilrate = d['tbilrate'][:-1] endogg = gs_l_realinv exogg = add_constant(np.c_[gs_l_realgdp, lint]) exogg2 = add_constant(np.c_[gs_l_realgdp, tbilrate]) res_ols = OLS(endogg, exogg).fit() res_ols2 = OLS(endogg, exogg2).fit() #the following were done accidentally with res_ols1 in R, #with original Greene data params = np.array([-272.3986041341653, 0.1779455206941112, 0.2149432424658157]) cov_hac_4 = np.array([1321.569466333051, -0.2318836566017612, 37.01280466875694, -0.2318836566017614, 4.602339488102263e-05, -0.0104687835998635, 37.012804668757, -0.0104687835998635, 21.16037144168061]).reshape(3,3, order='F') cov_hac_10 = np.array([2027.356101193361, -0.3507514463299015, 54.81079621448568, -0.350751446329901, 6.953380432635583e-05, -0.01268990195095196, 54.81079621448564, -0.01268990195095195, 22.92512402151113]).reshape(3,3, order='F') #goldfeld-quandt het_gq_greater = dict(statistic=13.20512768685082, df1=99, df2=98, pvalue=1.246141976112324e-30, distr='f') het_gq_less = dict(statistic=13.20512768685082, df1=99, df2=98, pvalue=1.) het_gq_2sided = dict(statistic=13.20512768685082, df1=99, df2=98, pvalue=1.246141976112324e-30, distr='f') #goldfeld-quandt, fraction = 0.5 het_gq_greater_2 = dict(statistic=87.1328934692124, df1=48, df2=47, pvalue=2.154956842194898e-33, distr='f') gq = smsdia.het_goldfeldquandt(endog, exog, split=0.5) compare_t_est(gq, het_gq_greater, decimal=(13, 14)) assert_equal(gq[-1], 'increasing') harvey_collier = dict(stat=2.28042114041313, df=199, pvalue=0.02364236161988260, distr='t') #hc = harvtest(fm, order.by=ggdp , data = list()) harvey_collier_2 = dict(stat=0.7516918462158783, df=199, pvalue=0.4531244858006127, distr='t') ################################## class TestDiagnosticG(object): @classmethod def setup_class(cls): d = macrodata.load_pandas().data #growth rates gs_l_realinv = 400 * np.diff(np.log(d['realinv'].values)) gs_l_realgdp = 400 * np.diff(np.log(d['realgdp'].values)) lint = d['realint'][:-1].values tbilrate = d['tbilrate'][:-1].values endogg = gs_l_realinv exogg = add_constant(np.c_[gs_l_realgdp, lint]) exogg2 = add_constant(np.c_[gs_l_realgdp, tbilrate]) exogg3 = add_constant(np.c_[gs_l_realgdp]) res_ols = OLS(endogg, exogg).fit() res_ols2 = OLS(endogg, exogg2).fit() res_ols3 = OLS(endogg, exogg3).fit() cls.res = res_ols cls.res2 = res_ols2 cls.res3 = res_ols3 cls.endog = cls.res.model.endog cls.exog = cls.res.model.exog def test_basic(self): #mainly to check I got the right regression #> mkarray(fm$coefficients, "params") params = np.array([-9.48167277465485, 4.3742216647032, -0.613996969478989]) assert_almost_equal(self.res.params, params, decimal=12) def test_hac(self): res = self.res #> nw = NeweyWest(fm, lag = 4, prewhite = FALSE, verbose=TRUE) #> nw2 = NeweyWest(fm, lag=10, prewhite = FALSE, verbose=TRUE) #> mkarray(nw, "cov_hac_4") cov_hac_4 = np.array([1.385551290884014, -0.3133096102522685, -0.0597207976835705, -0.3133096102522685, 0.1081011690351306, 0.000389440793564336, -0.0597207976835705, 0.000389440793564339, 0.0862118527405036]).reshape(3,3, order='F') #> mkarray(nw2, "cov_hac_10") cov_hac_10 = np.array([1.257386180080192, -0.2871560199899846, -0.03958300024627573, -0.2871560199899845, 0.1049107028987101, 0.0003896205316866944, -0.03958300024627578, 0.0003896205316866961, 0.0985539340694839]).reshape(3,3, order='F') cov = sw.cov_hac_simple(res, nlags=4, use_correction=False) bse_hac = sw.se_cov(cov) assert_almost_equal(cov, cov_hac_4, decimal=14) assert_almost_equal(bse_hac, np.sqrt(np.diag(cov)), decimal=14) cov = sw.cov_hac_simple(res, nlags=10, use_correction=False) bse_hac = sw.se_cov(cov) assert_almost_equal(cov, cov_hac_10, decimal=14) assert_almost_equal(bse_hac, np.sqrt(np.diag(cov)), decimal=14) def test_het_goldfeldquandt(self): #TODO: test options missing #> gq = gqtest(fm, alternative='greater') #> mkhtest_f(gq, 'het_gq_greater', 'f') het_gq_greater = dict(statistic=0.5313259064778423, pvalue=0.9990217851193723, parameters=(98, 98), distr='f') #> gq = gqtest(fm, alternative='less') #> mkhtest_f(gq, 'het_gq_less', 'f') het_gq_less = dict(statistic=0.5313259064778423, pvalue=0.000978214880627621, parameters=(98, 98), distr='f') #> gq = gqtest(fm, alternative='two.sided') #> mkhtest_f(gq, 'het_gq_two_sided', 'f') het_gq_two_sided = dict(statistic=0.5313259064778423, pvalue=0.001956429761255241, parameters=(98, 98), distr='f') #> gq = gqtest(fm, fraction=0.1, alternative='two.sided') #> mkhtest_f(gq, 'het_gq_two_sided_01', 'f') het_gq_two_sided_01 = dict(statistic=0.5006976835928314, pvalue=0.001387126702579789, parameters=(88, 87), distr='f') #> gq = gqtest(fm, fraction=0.5, alternative='two.sided') #> mkhtest_f(gq, 'het_gq_two_sided_05', 'f') het_gq_two_sided_05 = dict(statistic=0.434815645134117, pvalue=0.004799321242905568, parameters=(48, 47), distr='f') endogg, exogg = self.endog, self.exog #tests gq = smsdia.het_goldfeldquandt(endogg, exogg, split=0.5) compare_t_est(gq, het_gq_greater, decimal=(14, 14)) assert_equal(gq[-1], 'increasing') gq = smsdia.het_goldfeldquandt(endogg, exogg, split=0.5, alternative='decreasing') compare_t_est(gq, het_gq_less, decimal=(14, 14)) assert_equal(gq[-1], 'decreasing') gq = smsdia.het_goldfeldquandt(endogg, exogg, split=0.5, alternative='two-sided') compare_t_est(gq, het_gq_two_sided, decimal=(14, 14)) assert_equal(gq[-1], 'two-sided') #TODO: forcing the same split as R 202-90-90-1=21 gq = smsdia.het_goldfeldquandt(endogg, exogg, split=90, drop=21, alternative='two-sided') compare_t_est(gq, het_gq_two_sided_01, decimal=(14, 14)) assert_equal(gq[-1], 'two-sided') #TODO other options ??? def test_het_breusch_pagan(self): res = self.res bptest = dict(statistic=0.709924388395087, pvalue=0.701199952134347, parameters=(2,), distr='f') bp = smsdia.het_breuschpagan(res.resid, res.model.exog) compare_t_est(bp, bptest, decimal=(12, 12)) def test_het_white(self): res = self.res #TODO: regressiontest, compare with Greene or Gretl or Stata hw = smsdia.het_white(res.resid, res.model.exog) hw_values = (33.503722896538441, 2.9887960597830259e-06, 7.7945101228430946, 1.0354575277704231e-06) assert_almost_equal(hw, hw_values) def test_het_arch(self): #test het_arch and indirectly het_lm against R #> library(FinTS) #> at = ArchTest(residuals(fm), lags=4) #> mkhtest(at, 'archtest_4', 'chi2') archtest_4 = dict(statistic=3.43473400836259, pvalue=0.487871315392619, parameters=(4,), distr='chi2') #> at = ArchTest(residuals(fm), lags=12) #> mkhtest(at, 'archtest_12', 'chi2') archtest_12 = dict(statistic=8.648320999014171, pvalue=0.732638635007718, parameters=(12,), distr='chi2') at4 = smsdia.het_arch(self.res.resid, maxlag=4) at12 = smsdia.het_arch(self.res.resid, maxlag=12) compare_t_est(at4[:2], archtest_4, decimal=(12, 13)) compare_t_est(at12[:2], archtest_12, decimal=(12, 13)) def test_het_arch2(self): #test autolag options, this also test het_lm #unfortunately optimal lag=1 for this data resid = self.res.resid res1 = smsdia.het_arch(resid, maxlag=1, autolag=None, store=True) rs1 = res1[-1] res2 = smsdia.het_arch(resid, maxlag=5, autolag='aic', store=True) rs2 = res2[-1] assert_almost_equal(rs2.resols.params, rs1.resols.params, decimal=13) assert_almost_equal(res2[:4], res1[:4], decimal=13) #test that smallest lag, maxlag=1 works res3 = smsdia.het_arch(resid, maxlag=1, autolag='aic') assert_almost_equal(res3[:4], res1[:4], decimal=13) def test_acorr_breusch_godfrey(self): res = self.res #bgf = bgtest(fm, order = 4, type="F") breuschgodfrey_f = dict(statistic=1.179280833676792, pvalue=0.321197487261203, parameters=(4,195,), distr='f') #> bgc = bgtest(fm, order = 4, type="Chisq") #> mkhtest(bgc, "breuschpagan_c", "chi2") breuschgodfrey_c = dict(statistic=4.771042651230007, pvalue=0.3116067133066697, parameters=(4,), distr='chi2') bg = smsdia.acorr_breusch_godfrey(res, nlags=4) bg_r = [breuschgodfrey_c['statistic'], breuschgodfrey_c['pvalue'], breuschgodfrey_f['statistic'], breuschgodfrey_f['pvalue']] assert_almost_equal(bg, bg_r, decimal=13) # check that lag choice works bg2 = smsdia.acorr_breusch_godfrey(res, nlags=None) bg3 = smsdia.acorr_breusch_godfrey(res, nlags=14) assert_almost_equal(bg2, bg3, decimal=13) def test_acorr_ljung_box(self): #unit-test which may be useful later #ddof correction for fitted parameters in ARMA(p,q) fitdf=p+q #> bt = Box.test(residuals(fm), lag=4, type = "Ljung-Box", fitdf=2) #> mkhtest(bt, "ljung_box_4df2", "chi2") # ljung_box_4df2 = dict(statistic=5.23587172795227, # pvalue=0.0729532930400377, # parameters=(2,), distr='chi2') #> bt = Box.test(residuals(fm), lag=4, type = "Box-Pierce", fitdf=2) #> mkhtest(bt, "ljung_box_bp_4df2", "chi2") # ljung_box_bp_4df2 = dict(statistic=5.12462932741681, # pvalue=0.0771260128929921, # parameters=(2,), distr='chi2') res = self.res #general test #> bt = Box.test(residuals(fm), lag=4, type = "Ljung-Box") #> mkhtest(bt, "ljung_box_4", "chi2") ljung_box_4 = dict(statistic=5.23587172795227, pvalue=0.263940335284713, parameters=(4,), distr='chi2') #> bt = Box.test(residuals(fm), lag=4, type = "Box-Pierce") #> mkhtest(bt, "ljung_box_bp_4", "chi2") ljung_box_bp_4 = dict(statistic=5.12462932741681, pvalue=0.2747471266820692, parameters=(4,), distr='chi2') lb, lbpval, bp, bppval = smsdia.acorr_ljungbox(res.resid, 4, boxpierce=True) compare_t_est([lb[-1], lbpval[-1]], ljung_box_4, decimal=(13, 13)) compare_t_est([bp[-1], bppval[-1]], ljung_box_bp_4, decimal=(13, 13)) def test_acorr_ljung_box_big_default(self): res = self.res #test with big dataset and default lag #> bt = Box.test(residuals(fm), type = "Ljung-Box") #> mkhtest(bt, "ljung_box_none", "chi2") ljung_box_none = dict(statistic=51.03724531797195, pvalue=0.11334744923390, distr='chi2') #> bt = Box.test(residuals(fm), type = "Box-Pierce") #> mkhtest(bt, "ljung_box_bp_none", "chi2") ljung_box_bp_none = dict(statistic=45.12238537034000, pvalue=0.26638168491464, distr='chi2') lb, lbpval, bp, bppval = smsdia.acorr_ljungbox(res.resid, boxpierce=True) compare_t_est([lb[-1], lbpval[-1]], ljung_box_none, decimal=(13, 13)) compare_t_est([bp[-1], bppval[-1]], ljung_box_bp_none, decimal=(13, 13)) def test_acorr_ljung_box_small_default(self): res = self.res #test with small dataset and default lag #> bt = Box.test(residuals(fm), type = "Ljung-Box") #> mkhtest(bt, "ljung_box_small", "chi2") ljung_box_small = dict(statistic=9.61503968281915, pvalue=0.72507000996945, parameters=(0,), distr='chi2') #> bt = Box.test(residuals(fm), type = "Box-Pierce") #> mkhtest(bt, "ljung_box_bp_small", "chi2") ljung_box_bp_small = dict(statistic=7.41692150864936, pvalue=0.87940785887006, parameters=(0,), distr='chi2') lb, lbpval, bp, bppval = smsdia.acorr_ljungbox(res.resid[:30], boxpierce=True) compare_t_est([lb[-1], lbpval[-1]], ljung_box_small, decimal=(13, 13)) compare_t_est([bp[-1], bppval[-1]], ljung_box_bp_small, decimal=(13, 13)) def test_harvey_collier(self): #> hc = harvtest(fm, order.by = NULL, data = list()) #> mkhtest_f(hc, 'harvey_collier', 't') harvey_collier = dict(statistic=0.494432160939874, pvalue=0.6215491310408242, parameters=(198), distr='t') #> hc2 = harvtest(fm, order.by=ggdp , data = list()) #> mkhtest_f(hc2, 'harvey_collier_2', 't') harvey_collier_2 = dict(statistic=1.42104628340473, pvalue=0.1568762892441689, parameters=(198), distr='t') hc = smsdia.linear_harvey_collier(self.res) compare_t_est(hc, harvey_collier, decimal=(12, 12)) def test_rainbow(self): #rainbow test #> rt = raintest(fm) #> mkhtest_f(rt, 'raintest', 'f') raintest = dict(statistic=0.6809600116739604, pvalue=0.971832843583418, parameters=(101, 98), distr='f') #> rt = raintest(fm, center=0.4) #> mkhtest_f(rt, 'raintest_center_04', 'f') raintest_center_04 = dict(statistic=0.682635074191527, pvalue=0.971040230422121, parameters=(101, 98), distr='f') #> rt = raintest(fm, fraction=0.4) #> mkhtest_f(rt, 'raintest_fraction_04', 'f') raintest_fraction_04 = dict(statistic=0.565551237772662, pvalue=0.997592305968473, parameters=(122, 77), distr='f') #> rt = raintest(fm, order.by=ggdp) #Warning message: #In if (order.by == "mahalanobis") { : # the condition has length > 1 and only the first element will be used #> mkhtest_f(rt, 'raintest_order_gdp', 'f') raintest_order_gdp = dict(statistic=1.749346160513353, pvalue=0.002896131042494884, parameters=(101, 98), distr='f') rb = smsdia.linear_rainbow(self.res) compare_t_est(rb, raintest, decimal=(13, 14)) rb = smsdia.linear_rainbow(self.res, frac=0.4) compare_t_est(rb, raintest_fraction_04, decimal=(13, 14)) def test_compare_lr(self): res = self.res res3 = self.res3 #nested within res #lrtest #lrt = lrtest(fm, fm2) #Model 1: ginv ~ ggdp + lint #Model 2: ginv ~ ggdp lrtest = dict(loglike1=-763.9752181602237, loglike2=-766.3091902020184, chi2value=4.66794408358942, pvalue=0.03073069384028677, df=(4,3,1)) lrt = res.compare_lr_test(res3) assert_almost_equal(lrt[0], lrtest['chi2value'], decimal=11) assert_almost_equal(lrt[1], lrtest['pvalue'], decimal=11) waldtest = dict(fvalue=4.65216373312492, pvalue=0.03221346195239025, df=(199,200,1)) wt = res.compare_f_test(res3) assert_almost_equal(wt[0], waldtest['fvalue'], decimal=11) assert_almost_equal(wt[1], waldtest['pvalue'], decimal=11) def test_compare_nonnested(self): res = self.res res2 = self.res2 #jt = jtest(fm, lm(ginv ~ ggdp + tbilrate)) #Estimate Std. Error t value Pr(>|t|) jtest = [('M1 + fitted(M2)', 1.591505670785873, 0.7384552861695823, 2.155182176352370, 0.032354572525314450, '*'), ('M2 + fitted(M1)', 1.305687653016899, 0.4808385176653064, 2.715438978051544, 0.007203854534057954, '**')] jt1 = smsdia.compare_j(res2, res) assert_almost_equal(jt1, jtest[0][3:5], decimal=13) jt2 = smsdia.compare_j(res, res2) assert_almost_equal(jt2, jtest[1][3:5], decimal=14) #Estimate Std. Error z value Pr(>|z|) coxtest = [('fitted(M1) ~ M2', -0.782030488930356, 0.599696502782265, -1.304043770977755, 1.922186587840554e-01, ' '), ('fitted(M2) ~ M1', -2.248817107408537, 0.392656854330139, -5.727181590258883, 1.021128495098556e-08, '***')] ct1 = smsdia.compare_cox(res, res2) assert_almost_equal(ct1, coxtest[0][3:5], decimal=13) ct2 = smsdia.compare_cox(res2, res) assert_almost_equal(ct2, coxtest[1][3:5], decimal=12) #TODO should be approx # Res.Df Df F Pr(>F) encomptest = [('M1 vs. ME', 198, -1, 4.644810213266983, 0.032354572525313666, '*'), ('M2 vs. ME', 198, -1, 7.373608843521585, 0.007203854534058054, '**')] # Estimate Std. Error t value petest = [('M1 + log(fit(M1))-fit(M2)', -229.281878354594596, 44.5087822087058598, -5.15139, 6.201281252449979e-07), ('M2 + fit(M1)-exp(fit(M2))', 0.000634664704814, 0.0000462387010349, 13.72583, 1.319536115230356e-30)] def test_cusum_ols(self): #R library(strucchange) #> sc = sctest(ginv ~ ggdp + lint, type="OLS-CUSUM") #> mkhtest(sc, 'cusum_ols', 'BB') cusum_ols = dict(statistic=1.055750610401214, pvalue=0.2149567397376543, parameters=(), distr='BB') #Brownian Bridge k_vars=3 cs_ols = smsdia.breaks_cusumolsresid(self.res.resid, ddof=k_vars) # compare_t_est(cs_ols, cusum_ols, decimal=(12, 12)) def test_breaks_hansen(self): #> sc = sctest(ginv ~ ggdp + lint, type="Nyblom-Hansen") #> mkhtest(sc, 'breaks_nyblom_hansen', 'BB') breaks_nyblom_hansen = dict(statistic=1.0300792740544484, pvalue=0.1136087530212015, parameters=(), distr='BB') bh = smsdia.breaks_hansen(self.res) assert_almost_equal(bh[0], breaks_nyblom_hansen['statistic'], decimal=13) #TODO: breaks_hansen doesn't return pvalues def test_recursive_residuals(self): reccumres_standardize = np.array([-2.151, -3.748, -3.114, -3.096, -1.865, -2.230, -1.194, -3.500, -3.638, -4.447, -4.602, -4.631, -3.999, -4.830, -5.429, -5.435, -6.554, -8.093, -8.567, -7.532, -7.079, -8.468, -9.320, -12.256, -11.932, -11.454, -11.690, -11.318, -12.665, -12.842, -11.693, -10.803, -12.113, -12.109, -13.002, -11.897, -10.787, -10.159, -9.038, -9.007, -8.634, -7.552, -7.153, -6.447, -5.183, -3.794, -3.511, -3.979, -3.236, -3.793, -3.699, -5.056, -5.724, -4.888, -4.309, -3.688, -3.918, -3.735, -3.452, -2.086, -6.520, -7.959, -6.760, -6.855, -6.032, -4.405, -4.123, -4.075, -3.235, -3.115, -3.131, -2.986, -1.813, -4.824, -4.424, -4.796, -4.000, -3.390, -4.485, -4.669, -4.560, -3.834, -5.507, -3.792, -2.427, -1.756, -0.354, 1.150, 0.586, 0.643, 1.773, -0.830, -0.388, 0.517, 0.819, 2.240, 3.791, 3.187, 3.409, 2.431, 0.668, 0.957, -0.928, 0.327, -0.285, -0.625, -2.316, -1.986, -0.744, -1.396, -1.728, -0.646, -2.602, -2.741, -2.289, -2.897, -1.934, -2.532, -3.175, -2.806, -3.099, -2.658, -2.487, -2.515, -2.224, -2.416, -1.141, 0.650, -0.947, 0.725, 0.439, 0.885, 2.419, 2.642, 2.745, 3.506, 4.491, 5.377, 4.624, 5.523, 6.488, 6.097, 5.390, 6.299, 6.656, 6.735, 8.151, 7.260, 7.846, 8.771, 8.400, 8.717, 9.916, 9.008, 8.910, 8.294, 8.982, 8.540, 8.395, 7.782, 7.794, 8.142, 8.362, 8.400, 7.850, 7.643, 8.228, 6.408, 7.218, 7.699, 7.895, 8.725, 8.938, 8.781, 8.350, 9.136, 9.056, 10.365, 10.495, 10.704, 10.784, 10.275, 10.389, 11.586, 11.033, 11.335, 11.661, 10.522, 10.392, 10.521, 10.126, 9.428, 9.734, 8.954, 9.949, 10.595, 8.016, 6.636, 6.975]) rr = smsdia.recursive_olsresiduals(self.res, skip=3, alpha=0.95) assert_equal(np.round(rr[5][1:], 3), reccumres_standardize) #extra zero in front #assert_equal(np.round(rr[3][4:], 3), np.diff(reccumres_standardize)) assert_almost_equal(rr[3][4:], np.diff(reccumres_standardize),3) assert_almost_equal(rr[4][3:].std(ddof=1), 10.7242, decimal=4) #regression number, visually checked with graph from gretl ub0 = np.array([ 13.37318571, 13.50758959, 13.64199346, 13.77639734, 13.91080121]) ub1 = np.array([ 39.44753774, 39.58194162, 39.7163455 , 39.85074937, 39.98515325]) lb, ub = rr[6] assert_almost_equal(ub[:5], ub0, decimal=7) assert_almost_equal(lb[:5], -ub0, decimal=7) assert_almost_equal(ub[-5:], ub1, decimal=7) assert_almost_equal(lb[-5:], -ub1, decimal=7) #test a few values with explicit OLS endog = self.res.model.endog exog = self.res.model.exog params = [] ypred = [] for i in range(3,10): resi = OLS(endog[:i], exog[:i]).fit() ypred.append(resi.model.predict(resi.params, exog[i])) params.append(resi.params) assert_almost_equal(rr[2][3:10], ypred, decimal=12) assert_almost_equal(rr[0][3:10], endog[3:10] - ypred, decimal=12) assert_almost_equal(rr[1][2:9], params, decimal=12) def test_normality(self): res = self.res #> library(nortest) #Lilliefors (Kolmogorov-Smirnov) normality test #> lt = lillie.test(residuals(fm)) #> mkhtest(lt, "lilliefors", "-") lilliefors1 = dict(statistic=0.0723390908786589, pvalue=0.01204113540102896, parameters=(), distr='-') #> lt = lillie.test(residuals(fm)**2) #> mkhtest(lt, "lilliefors", "-") lilliefors2 = dict(statistic=0.301311621898024, pvalue=1.004305736618051e-51, parameters=(), distr='-') #> lt = lillie.test(residuals(fm)[1:20]) #> mkhtest(lt, "lilliefors", "-") lilliefors3 = dict(statistic=0.1333956004203103, pvalue=0.20, parameters=(), distr='-') lf1 = smsdia.lilliefors(res.resid) lf2 = smsdia.lilliefors(res.resid**2) lf3 = smsdia.lilliefors(res.resid[:20]) compare_t_est(lf1, lilliefors1, decimal=(14, 14)) compare_t_est(lf2, lilliefors2, decimal=(14, 14)) # pvalue very small assert_allclose(lf2[1], lilliefors2['pvalue'], rtol=1e-10) compare_t_est(lf3, lilliefors3, decimal=(14, 1)) # R uses different approximation for pvalue in last case #> ad = ad.test(residuals(fm)) #> mkhtest(ad, "ad3", "-") adr1 = dict(statistic=1.602209621518313, pvalue=0.0003937979149362316, parameters=(), distr='-') #> ad = ad.test(residuals(fm)**2) #> mkhtest(ad, "ad3", "-") adr2 = dict(statistic=np.inf, pvalue=np.nan, parameters=(), distr='-') #> ad = ad.test(residuals(fm)[1:20]) #> mkhtest(ad, "ad3", "-") adr3 = dict(statistic=0.3017073732210775, pvalue=0.5443499281265933, parameters=(), distr='-') ad1 = smsdia.normal_ad(res.resid) compare_t_est(ad1, adr1, decimal=(11, 13)) ad2 = smsdia.normal_ad(res.resid**2) assert_(np.isinf(ad2[0])) ad3 = smsdia.normal_ad(res.resid[:20]) compare_t_est(ad3, adr3, decimal=(11, 12)) def test_influence(self): res = self.res #this test is slow infl = oi.OLSInfluence(res) path = os.path.join(cur_dir, "results", "influence_lsdiag_R.json") with open(path, 'r') as fp: lsdiag = json.load(fp) #basic assert_almost_equal(np.array(lsdiag['cov.scaled']).reshape(3, 3), res.cov_params(), decimal=14) assert_almost_equal(np.array(lsdiag['cov.unscaled']).reshape(3, 3), res.normalized_cov_params, decimal=14) c0, c1 = infl.cooks_distance #TODO: what's c1 assert_almost_equal(c0, lsdiag['cooks'], decimal=14) assert_almost_equal(infl.hat_matrix_diag, lsdiag['hat'], decimal=14) assert_almost_equal(infl.resid_studentized_internal, lsdiag['std.res'], decimal=14) #slow: #infl._get_all_obs() #slow, nobs estimation loop, called implicitly dffits, dffth = infl.dffits assert_almost_equal(dffits, lsdiag['dfits'], decimal=14) assert_almost_equal(infl.resid_studentized_external, lsdiag['stud.res'], decimal=14) import pandas fn = os.path.join(cur_dir,"results/influence_measures_R.csv") infl_r = pandas.read_csv(fn, index_col=0) conv = lambda s: 1 if s=='TRUE' else 0 fn = os.path.join(cur_dir,"results/influence_measures_bool_R.csv") #not used yet: #infl_bool_r = pandas.read_csv(fn, index_col=0, # converters=dict(zip(lrange(7),[conv]*7))) infl_r2 = np.asarray(infl_r) assert_almost_equal(infl.dfbetas, infl_r2[:,:3], decimal=13) assert_almost_equal(infl.cov_ratio, infl_r2[:,4], decimal=14) #duplicates assert_almost_equal(dffits, infl_r2[:,3], decimal=14) assert_almost_equal(c0, infl_r2[:,5], decimal=14) assert_almost_equal(infl.hat_matrix_diag, infl_r2[:,6], decimal=14) #Note: for dffits, R uses a threshold around 0.36, mine: dffits[1]=0.24373 #TODO: finish and check thresholds and pvalues ''' R has >>> np.nonzero(np.asarray(infl_bool_r["dffit"]))[0] array([ 6, 26, 63, 76, 90, 199]) >>> np.nonzero(np.asarray(infl_bool_r["cov.r"]))[0] array([ 4, 26, 59, 61, 63, 72, 76, 84, 91, 92, 94, 95, 108, 197, 198]) >>> np.nonzero(np.asarray(infl_bool_r["hat"]))[0] array([ 62, 76, 84, 90, 91, 92, 95, 108, 197, 199]) ''' class TestDiagnosticGPandas(TestDiagnosticG): @classmethod def setup_class(cls): d = macrodata.load_pandas().data #growth rates d['gs_l_realinv'] = 400 * np.log(d['realinv']).diff() d['gs_l_realgdp'] = 400 * np.log(d['realgdp']).diff() d['lint'] = d['realint'].shift(1) d['tbilrate'] = d['tbilrate'].shift(1) d = d.dropna() cls.d = d endogg = d['gs_l_realinv'] exogg = add_constant(d[['gs_l_realgdp', 'lint']]) exogg2 = add_constant(d[['gs_l_realgdp', 'tbilrate']]) exogg3 = add_constant(d[['gs_l_realgdp']]) res_ols = OLS(endogg, exogg).fit() res_ols2 = OLS(endogg, exogg2).fit() res_ols3 = OLS(endogg, exogg3).fit() cls.res = res_ols cls.res2 = res_ols2 cls.res3 = res_ols3 cls.endog = cls.res.model.endog cls.exog = cls.res.model.exog def test_spec_white(): resdir = os.path.join(cur_dir, "results") wsfiles = ['wspec1.csv', 'wspec2.csv', 'wspec3.csv', 'wspec4.csv'] for file in wsfiles: mdlfile = os.path.join(resdir, file) mdl = np.asarray(pd.read_csv(mdlfile)) # DV is in last column lastcol = mdl.shape[1] - 1 dv = mdl[:,lastcol] # create design matrix design = np.concatenate((np.ones((mdl.shape[0], 1)), \ np.delete(mdl, lastcol, 1)), axis=1) # perform OLS and generate residuals resids = dv - np.dot(design, np.linalg.lstsq(design, dv, rcond=-1)[0]) # perform White spec test. wspec3/wspec4 contain dummies. wsres = smsdia.spec_white(resids, design) # compare results to SAS 9.3 output if file == 'wspec1.csv': assert_almost_equal(wsres, [3.251, 0.661, 5], decimal=3) elif file == 'wspec2.csv': assert_almost_equal(wsres, [6.070, 0.733, 9], decimal=3) elif file == 'wspec3.csv': assert_almost_equal(wsres, [6.767, 0.454, 7], decimal=3) else: assert_almost_equal(wsres, [8.462, 0.671, 11], decimal=3) def grangertest(): #> gt = grangertest(ginv, ggdp, order=4) #> gt #Granger causality test # #Model 1: ggdp ~ Lags(ggdp, 1:4) + Lags(ginv, 1:4) #Model 2: ggdp ~ Lags(ggdp, 1:4) grangertest = dict(fvalue=1.589672703015157, pvalue=0.178717196987075, df=(198,193)) @pytest.mark.smoke def test_outlier_influence_funcs(reset_randomstate): x = add_constant(np.random.randn(10, 2)) y = x.sum(1) + np.random.randn(10) res = OLS(y, x).fit() out_05 = oi.summary_table(res) # GH3344 : Check alpha has an effect out_01 = oi.summary_table(res, alpha=0.01) assert_(np.all(out_01[1][:, 6] <= out_05[1][:, 6])) assert_(np.all(out_01[1][:, 7] >= out_05[1][:, 7])) res2 = OLS(y, x[:,0]).fit() oi.summary_table(res2, alpha=0.05) infl = res2.get_influence() infl.summary_table() def test_influence_wrapped(): from pandas import DataFrame d = macrodata.load_pandas().data #growth rates gs_l_realinv = 400 * np.log(d['realinv']).diff().dropna() gs_l_realgdp = 400 * np.log(d['realgdp']).diff().dropna() lint = d['realint'][:-1] # re-index these because they won't conform to lint gs_l_realgdp.index = lint.index gs_l_realinv.index = lint.index data = dict(const=np.ones_like(lint), lint=lint, lrealgdp=gs_l_realgdp) #order is important exog = DataFrame(data, columns=['const','lrealgdp','lint']) res = OLS(gs_l_realinv, exog).fit() #basic # already tested #assert_almost_equal(lsdiag['cov.scaled'], # res.cov_params().values.ravel(), decimal=14) #assert_almost_equal(lsdiag['cov.unscaled'], # res.normalized_cov_params.values.ravel(), decimal=14) infl = oi.OLSInfluence(res) # smoke test just to make sure it works, results separately tested df = infl.summary_frame() assert_(isinstance(df, DataFrame)) #this test is slow path = os.path.join(cur_dir, "results", "influence_lsdiag_R.json") with open(path, "r") as fp: lsdiag = json.load(fp) c0, c1 = infl.cooks_distance #TODO: what's c1, it's pvalues? -ss #NOTE: we get a hard-cored 5 decimals with pandas testing assert_almost_equal(c0, lsdiag['cooks'], 14) assert_almost_equal(infl.hat_matrix_diag, (lsdiag['hat']), 14) assert_almost_equal(infl.resid_studentized_internal, lsdiag['std.res'], 14) #slow: dffits, dffth = infl.dffits assert_almost_equal(dffits, lsdiag['dfits'], 14) assert_almost_equal(infl.resid_studentized_external, lsdiag['stud.res'], 14) import pandas fn = os.path.join(cur_dir,"results/influence_measures_R.csv") infl_r = pandas.read_csv(fn, index_col=0) conv = lambda s: 1 if s=='TRUE' else 0 fn = os.path.join(cur_dir,"results/influence_measures_bool_R.csv") #not used yet: #infl_bool_r = pandas.read_csv(fn, index_col=0, # converters=dict(zip(lrange(7),[conv]*7))) infl_r2 = np.asarray(infl_r) #TODO: finish wrapping this stuff assert_almost_equal(infl.dfbetas, infl_r2[:,:3], decimal=13) assert_almost_equal(infl.cov_ratio, infl_r2[:,4], decimal=14) def test_influence_dtype(): # see #2148 bug when endog is integer y = np.ones(20) np.random.seed(123) x = np.random.randn(20, 3) res1 = OLS(y, x).fit() res2 = OLS(y*1., x).fit() cr1 = res1.get_influence().cov_ratio cr2 = res2.get_influence().cov_ratio assert_allclose(cr1, cr2, rtol=1e-14) # regression test for values cr3 = np.array( [ 1.22239215, 1.31551021, 1.52671069, 1.05003921, 0.89099323, 1.57405066, 1.03230092, 0.95844196, 1.15531836, 1.21963623, 0.87699564, 1.16707748, 1.10481391, 0.98839447, 1.08999334, 1.35680102, 1.46227715, 1.45966708, 1.13659521, 1.22799038]) assert_almost_equal(cr1, cr3, decimal=8) def get_duncan_data(): # results from R with NA -> 1. Just testing interface here because # outlier_test is just a wrapper labels = ['accountant', 'pilot', 'architect', 'author', 'chemist', 'minister', 'professor', 'dentist', 'reporter', 'engineer', 'undertaker', 'lawyer', 'physician', 'welfare.worker', 'teacher', 'conductor', 'contractor', 'factory.owner', 'store.manager', 'banker', 'bookkeeper', 'mail.carrier', 'insurance.agent', 'store.clerk', 'carpenter', 'electrician', 'RR.engineer', 'machinist', 'auto.repairman', 'plumber', 'gas.stn.attendant', 'coal.miner', 'streetcar.motorman', 'taxi.driver', 'truck.driver', 'machine.operator', 'barber', 'bartender', 'shoe.shiner', 'cook', 'soda.clerk', 'watchman', 'janitor', 'policeman', 'waiter'] #Duncan's prestige data from car exog = [[1.0, 62.0, 86.0], [1.0, 72.0, 76.0], [1.0, 75.0, 92.0], [1.0, 55.0, 90.0], [1.0, 64.0, 86.0], [1.0, 21.0, 84.0], [1.0, 64.0, 93.0], [1.0, 80.0, 100.0], [1.0, 67.0, 87.0], [1.0, 72.0, 86.0], [1.0, 42.0, 74.0], [1.0, 76.0, 98.0], [1.0, 76.0, 97.0], [1.0, 41.0, 84.0], [1.0, 48.0, 91.0], [1.0, 76.0, 34.0], [1.0, 53.0, 45.0], [1.0, 60.0, 56.0], [1.0, 42.0, 44.0], [1.0, 78.0, 82.0], [1.0, 29.0, 72.0], [1.0, 48.0, 55.0], [1.0, 55.0, 71.0], [1.0, 29.0, 50.0], [1.0, 21.0, 23.0], [1.0, 47.0, 39.0], [1.0, 81.0, 28.0], [1.0, 36.0, 32.0], [1.0, 22.0, 22.0], [1.0, 44.0, 25.0], [1.0, 15.0, 29.0], [1.0, 7.0, 7.0], [1.0, 42.0, 26.0], [1.0, 9.0, 19.0], [1.0, 21.0, 15.0], [1.0, 21.0, 20.0], [1.0, 16.0, 26.0], [1.0, 16.0, 28.0], [1.0, 9.0, 17.0], [1.0, 14.0, 22.0], [1.0, 12.0, 30.0], [1.0, 17.0, 25.0], [1.0, 7.0, 20.0], [1.0, 34.0, 47.0], [1.0, 8.0, 32.0]] endog = [ 82., 83., 90., 76., 90., 87., 93., 90., 52., 88., 57., 89., 97., 59., 73., 38., 76., 81., 45., 92., 39., 34., 41., 16., 33., 53., 67., 57., 26., 29., 10., 15., 19., 10., 13., 24., 20., 7., 3., 16., 6., 11., 8., 41., 10.] return endog, exog, labels def test_outlier_test(): endog, exog, labels = get_duncan_data() ndarray_mod = OLS(endog, exog).fit() rstudent = [3.1345185839, -2.3970223990, 2.0438046359, -1.9309187757, 1.8870465798, -1.7604905300, -1.7040324156, 1.6024285876, -1.4332485037, -1.1044851583, 1.0688582315, 1.0185271840, -0.9024219332, -0.9023876471, -0.8830953936, 0.8265782334, 0.8089220547, 0.7682770197, 0.7319491074, -0.6665962829, 0.5227352794, -0.5135016547, 0.5083881518, 0.4999224372, -0.4980818221, -0.4759717075, -0.4293565820, -0.4114056499, -0.3779540862, 0.3556874030, 0.3409200462, 0.3062248646, 0.3038999429, -0.3030815773, -0.1873387893, 0.1738050251, 0.1424246593, -0.1292266025, 0.1272066463, -0.0798902878, 0.0788467222, 0.0722556991, 0.0505098280, 0.0233215136, 0.0007112055] unadj_p = [0.003177202, 0.021170298, 0.047432955, 0.060427645, 0.066248120, 0.085783008, 0.095943909, 0.116738318, 0.159368890, 0.275822623, 0.291386358, 0.314400295, 0.372104049, 0.372122040, 0.382333561, 0.413260793, 0.423229432, 0.446725370, 0.468363101, 0.508764039, 0.603971990, 0.610356737, 0.613905871, 0.619802317, 0.621087703, 0.636621083, 0.669911674, 0.682917818, 0.707414459, 0.723898263, 0.734904667, 0.760983108, 0.762741124, 0.763360242, 0.852319039, 0.862874018, 0.887442197, 0.897810225, 0.899398691, 0.936713197, 0.937538115, 0.942749758, 0.959961394, 0.981506948, 0.999435989] bonf_p = [0.1429741, 0.9526634, 2.1344830, 2.7192440, 2.9811654, 3.8602354, 4.3174759, 5.2532243, 7.1716001, 12.4120180, 13.1123861, 14.1480133, 16.7446822, 16.7454918, 17.2050103, 18.5967357, 19.0453245, 20.1026416, 21.0763395, 22.8943818, 27.1787396, 27.4660532, 27.6257642, 27.8911043, 27.9489466, 28.6479487, 30.1460253, 30.7313018, 31.8336506, 32.5754218, 33.0707100, 34.2442399, 34.3233506, 34.3512109, 38.3543568, 38.8293308, 39.9348989, 40.4014601, 40.4729411, 42.1520939, 42.1892152, 42.4237391, 43.1982627, 44.1678127, 44.9746195] bonf_p = np.array(bonf_p) bonf_p[bonf_p > 1] = 1 sorted_labels = ["minister", "reporter", "contractor", "insurance.agent", "machinist", "store.clerk", "conductor", "factory.owner", "mail.carrier", "streetcar.motorman", "carpenter", "coal.miner", "bartender", "bookkeeper", "soda.clerk", "chemist", "RR.engineer", "professor", "electrician", "gas.stn.attendant", "auto.repairman", "watchman", "banker", "machine.operator", "dentist", "waiter", "shoe.shiner", "welfare.worker", "plumber", "physician", "pilot", "engineer", "accountant", "lawyer", "undertaker", "barber", "store.manager", "truck.driver", "cook", "janitor", "policeman", "architect", "teacher", "taxi.driver", "author"] res2 = np.c_[rstudent, unadj_p, bonf_p] res = oi.outlier_test(ndarray_mod, method='b', labels=labels, order=True) np.testing.assert_almost_equal(res.values, res2, 7) np.testing.assert_equal(res.index.tolist(), sorted_labels) # pylint: disable-msg=E1103 data = pd.DataFrame(np.column_stack((endog, exog)), columns='y const var1 var2'.split(), index=labels) # check `order` with pandas bug in #3971 res_pd = OLS.from_formula('y ~ const + var1 + var2 - 0', data).fit() res_outl2 = oi.outlier_test(res_pd, method='b', order=True) assert_almost_equal(res_outl2.values, res2, 7) assert_equal(res_outl2.index.tolist(), sorted_labels) res_outl1 = res_pd.outlier_test(method='b') res_outl1 = res_outl1.sort_values(['unadj_p'], ascending=True) assert_almost_equal(res_outl1.values, res2, 7) assert_equal(res_outl1.index.tolist(), sorted_labels) assert_array_equal(res_outl2.index, res_outl1.index) # additional keywords in method res_outl3 = res_pd.outlier_test(method='b', order=True) assert_equal(res_outl3.index.tolist(), sorted_labels) res_outl4 = res_pd.outlier_test(method='b', order=True, cutoff=0.15) assert_equal(res_outl4.index.tolist(), sorted_labels[:1]) if __name__ == '__main__': import pytest pytest.main([__file__, '-vvs', '-x', '--pdb']) #t = TestDiagnosticG() #t.test_basic() #t.test_hac() #t.test_acorr_breusch_godfrey() #t.test_acorr_ljung_box() #t.test_het_goldfeldquandt() #t.test_het_breusch_pagan() #t.test_het_white() #t.test_compare_lr() #t.test_compare_nonnested() #t.test_influence() ################################################## ''' J test Model 1: ginv ~ ggdp + lint Model 2: ginv ~ ggdp + tbilrate Estimate Std. Error t value Pr(>|t|) M1 + fitted(M2) 1.591505670785873 0.7384552861695823 2.15518 0.0323546 * M2 + fitted(M1) 1.305687653016899 0.4808385176653064 2.71544 0.0072039 ** --- Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 = lm(ginv ~ ggdp + tbilrate) > ct = coxtest(fm, fm3) > ct Cox test Model 1: ginv ~ ggdp + lint Model 2: ginv ~ ggdp + tbilrate Estimate Std. Error z value Pr(>|z|) fitted(M1) ~ M2 -0.782030488930356 0.599696502782265 -1.30404 0.19222 fitted(M2) ~ M1 -2.248817107408537 0.392656854330139 -5.72718 1.0211e-08 *** --- Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 > et = encomptest(fm, fm3) > et Encompassing test Model 1: ginv ~ ggdp + lint Model 2: ginv ~ ggdp + tbilrate Model E: ginv ~ ggdp + lint + tbilrate Res.Df Df F Pr(>F) M1 vs. ME 198 -1 4.64481 0.0323546 * M2 vs. ME 198 -1 7.37361 0.0072039 ** --- Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 > fm4 = lm(realinv ~ realgdp + realint, data=d) > fm5 = lm(log(realinv) ~ realgdp + realint, data=d) > pet = petest(fm4, fm5) > pet PE test Model 1: realinv ~ realgdp + realint Model 2: log(realinv) ~ realgdp + realint Estimate Std. Error t value M1 + log(fit(M1))-fit(M2) -229.281878354594596 44.5087822087058598 -5.15139 M2 + fit(M1)-exp(fit(M2)) 0.000634664704814 0.0000462387010349 13.72583 Pr(>|t|) M1 + log(fit(M1))-fit(M2) 6.2013e-07 *** M2 + fit(M1)-exp(fit(M2)) < 2.22e-16 *** --- Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 '''