reconnect moved files to git repo
This commit is contained in:
root
Binary file not shown.
Binary file not shown.
Binary file not shown.
Binary file not shown.
@ -0,0 +1,39 @@
|
||||
food,income,persons
|
||||
15.998,62.476,1
|
||||
16.652,82.304,5
|
||||
21.741,74.679,3
|
||||
7.431,39.151,3
|
||||
10.481,64.724,5
|
||||
13.548,36.786,3
|
||||
23.256,83.052,4
|
||||
17.976,86.935,1
|
||||
14.161,88.233,2
|
||||
8.825,38.695,2
|
||||
14.184,73.831,7
|
||||
19.604,77.122,3
|
||||
13.728,45.519,2
|
||||
21.141,82.251,2
|
||||
17.446,59.862,3
|
||||
9.629,26.563,3
|
||||
14.005,61.818,2
|
||||
9.16,29.682,1
|
||||
18.831,50.825,5
|
||||
7.641,71.062,4
|
||||
13.882,41.99,4
|
||||
9.67,37.324,3
|
||||
21.604,86.352,5
|
||||
10.866,45.506,2
|
||||
28.98,69.929,6
|
||||
10.882,61.041,2
|
||||
18.561,82.469,1
|
||||
11.629,44.208,2
|
||||
18.067,49.467,5
|
||||
14.539,25.905,5
|
||||
19.192,79.178,5
|
||||
25.918,75.811,3
|
||||
28.833,82.718,6
|
||||
15.869,48.311,4
|
||||
14.91,42.494,5
|
||||
9.55,40.573,4
|
||||
23.066,44.872,6
|
||||
14.751,27.167,7
|
||||
|
@ -0,0 +1,37 @@
|
||||
,Unnamed: 0,gender,age,Basename,ID,id,plate,CpG,methylation
|
||||
0,0,F,58.231,6042324088_R04C01,age58.231_F,id_0,6042324088,CpG_0,0.815
|
||||
1,1,M,52.632,6042324088_R06C01,age52.632_M,id_1,6042324088,CpG_0,0.803
|
||||
2,2,M,64.679,6042324088_R01C01,age64.679_M,id_2,6042324088,CpG_0,0.803
|
||||
3,3,F,55.299,6042324088_R04C02,age55.299_F,id_3,6042324088,CpG_0,0.808
|
||||
4,4,M,56.019,6042324088_R02C01,age56.019_M,id_4,6042324088,CpG_0,0.8550000000000001
|
||||
5,5,M,62.021,6042324088_R01C02,age62.021_M,id_5,6042324088,CpG_0,0.8129999999999998
|
||||
6,6,F,52.298,6042324088_R06C02,age52.298_F,id_6,6042324088,CpG_0,0.816
|
||||
7,7,F,39.71,6042324088_R03C01,age39.71_F,id_7,6042324088,CpG_0,0.827
|
||||
8,8,F,57.492,6042324088_R05C02,age57.492_F,id_8,6042324088,CpG_0,0.829
|
||||
9,9,F,57.623999999999995,6042324088_R05C01,age57.624_F,id_9,6042324088,CpG_0,0.7760000000000001
|
||||
10,10,F,40.486999999999995,6042324088_R03C02,age40.487_F,id_10,6042324088,CpG_0,0.7859999999999999
|
||||
11,11,M,53.662,6042324088_R02C02,age53.662_M,id_11,6042324088,CpG_0,0.822
|
||||
12,12,F,58.231,6042324088_R04C01,age58.231_F,id_0,6042324088,CpG_1,0.891
|
||||
13,13,M,52.632,6042324088_R06C01,age52.632_M,id_1,6042324088,CpG_1,0.894
|
||||
14,14,M,64.679,6042324088_R01C01,age64.679_M,id_2,6042324088,CpG_1,0.894
|
||||
15,15,F,55.299,6042324088_R04C02,age55.299_F,id_3,6042324088,CpG_1,0.869
|
||||
16,16,M,56.019,6042324088_R02C01,age56.019_M,id_4,6042324088,CpG_1,0.914
|
||||
17,17,M,62.021,6042324088_R01C02,age62.021_M,id_5,6042324088,CpG_1,0.889
|
||||
18,18,F,52.298,6042324088_R06C02,age52.298_F,id_6,6042324088,CpG_1,0.8850000000000001
|
||||
19,19,F,39.71,6042324088_R03C01,age39.71_F,id_7,6042324088,CpG_1,0.898
|
||||
20,20,F,57.492,6042324088_R05C02,age57.492_F,id_8,6042324088,CpG_1,0.896
|
||||
21,21,F,57.623999999999995,6042324088_R05C01,age57.624_F,id_9,6042324088,CpG_1,0.86
|
||||
22,22,F,40.486999999999995,6042324088_R03C02,age40.487_F,id_10,6042324088,CpG_1,0.887
|
||||
23,23,M,53.662,6042324088_R02C02,age53.662_M,id_11,6042324088,CpG_1,0.8800000000000001
|
||||
24,24,F,58.231,6042324088_R04C01,age58.231_F,id_0,6042324088,CpG_2,0.936
|
||||
25,25,M,52.632,6042324088_R06C01,age52.632_M,id_1,6042324088,CpG_2,0.9129999999999999
|
||||
26,26,M,64.679,6042324088_R01C01,age64.679_M,id_2,6042324088,CpG_2,0.9000000000000001
|
||||
27,27,F,55.299,6042324088_R04C02,age55.299_F,id_3,6042324088,CpG_2,0.9119999999999999
|
||||
28,28,M,56.019,6042324088_R02C01,age56.019_M,id_4,6042324088,CpG_2,0.9349999999999999
|
||||
29,29,M,62.021,6042324088_R01C02,age62.021_M,id_5,6042324088,CpG_2,0.9280000000000002
|
||||
30,30,F,52.298,6042324088_R06C02,age52.298_F,id_6,6042324088,CpG_2,0.9150000000000001
|
||||
31,31,F,39.71,6042324088_R03C01,age39.71_F,id_7,6042324088,CpG_2,0.9160000000000001
|
||||
32,32,F,57.492,6042324088_R05C02,age57.492_F,id_8,6042324088,CpG_2,0.929
|
||||
33,33,F,57.623999999999995,6042324088_R05C01,age57.624_F,id_9,6042324088,CpG_2,0.92
|
||||
34,34,F,40.486999999999995,6042324088_R03C02,age40.487_F,id_10,6042324088,CpG_2,0.9160000000000001
|
||||
35,35,M,53.662,6042324088_R02C02,age53.662_M,id_11,6042324088,CpG_2,0.926
|
||||
|
@ -0,0 +1,37 @@
|
||||
"fittedvalues","pearson","deviance","response","weighted","sweighted","sweighted2"
|
||||
0.808801783033006,0.35152036037566,0.316277475404175,0.00619821696699407,0.0143437070668562,0.319573728733701,0.339140493550815
|
||||
0.819372283335466,-1.04578186069401,-1.05384293226473,-0.0163722833354659,-0.0424402211037504,-1.04203526956567,-1.12150619891079
|
||||
0.819372283335466,-0.848866164111184,-0.868303748623021,-0.0163722833354659,-0.0431126841865535,-0.858858035684054,-0.924509357407569
|
||||
0.808801783033006,-0.0478403593182754,-0.0740190330766135,-0.000801783033005821,-0.00346030425148218,-0.0811185264757917,-0.086082858454213
|
||||
0.819372283335466,2.14604894870802,2.23125643431909,0.0356277166645341,0.0991242044217954,2.29487706595179,2.46999256391156
|
||||
0.819372283335466,-0.345945859185448,-0.379745334132177,-0.00637228333546613,-0.0182042467717062,-0.379769622695936,-0.408782131481404
|
||||
0.808801783033006,0.452417702358772,0.422868818621262,0.00719821696699408,0.0172535750290291,0.426089629903614,0.452154192998679
|
||||
0.808801783033006,1.42260123423979,1.43173413510406,0.0181982169669941,0.0471933178688695,1.44999971504468,1.53857262907721
|
||||
0.808801783033006,1.16025896160801,1.15557258567386,0.0201982169669941,0.0519452178671859,1.17226143527927,1.24402734629728
|
||||
0.808801783033006,-1.87995098901712,-1.83681386957748,-0.0328017830330057,-0.0799212843345725,-1.79947817362265,-1.90964475994369
|
||||
0.808801783033006,-1.75862731269533,-1.73276624166304,-0.022801783033006,-0.0563377478383546,-1.70777742849756,-1.81210402283831
|
||||
0.819372283335466,0.164877426795495,0.125978175853291,0.00262771666453399,0.00542846196158609,0.130924012206529,0.140910550838054
|
||||
0.885906034257019,0.357336625125062,0.301293672955372,0.00509396574298104,0.0138352538385608,0.30824553489262,0.325442400805389
|
||||
0.892778886747752,0.096985881228438,0.0257403439859805,0.0012211132522485,0.00184829475110839,0.0453812037053184,0.0482934532444257
|
||||
0.892778886747752,0.0787239060703272,0.0625155549271173,0.0012211132522485,0.000749754320257284,0.0149360341368548,0.0158983762537736
|
||||
0.885906034257019,-1.24771994382442,-1.2462075883806,-0.016906034257019,-0.0520827370000992,-1.22095474074702,-1.28901416807562
|
||||
0.892778886747752,1.58942674175724,1.63373837694379,0.0212211132522485,0.07283953019518,1.68634662254971,1.79466886330034
|
||||
0.892778886747752,-0.255092478889325,-0.306807633696473,-0.00377888674775151,-0.0148849956658763,-0.31052475055691,-0.330511444147795
|
||||
0.885906034257019,-0.0704363879176134,-0.108791198935213,-0.000906034257018851,-0.00482003763451928,-0.119034347858826,-0.125664526445956
|
||||
0.885906034257019,1.16939404467676,1.16572330650475,0.012093965742981,0.0385653528018314,1.18490822638558,1.25074305918135
|
||||
0.885906034257019,0.71720205631466,0.678301944285756,0.010093965742981,0.0305419280332159,0.689247747193123,0.727692107851627
|
||||
0.885906034257019,-1.83648680424447,-1.7784513791609,-0.025906034257019,-0.0765923381476135,-1.72452484855445,-1.82071787157763
|
||||
0.885906034257019,0.104362993682885,0.0502517218075574,0.00109396574298104,0.00212749934649926,0.0644913135952729,0.068074968576542
|
||||
0.892778886747752,-0.997002863803822,-1.01348529078318,-0.0127788867477514,-0.0413363208029017,-0.996952176817101,-1.06094759772788
|
||||
0.918023532863447,1.4614350478071,1.49772318274252,0.0179764671365528,0.0697476840643581,1.55395863587393,1.63670046801479
|
||||
0.923129550226661,-0.934423619871682,-0.959684761138903,-0.0101295502266615,-0.0383660250384391,-0.942001483577578,-0.99713053437261
|
||||
0.923129550226661,-1.73188460406476,-1.66782538710199,-0.0231295502266613,-0.0802425599096661,-1.59853112100523,-1.69261746277774
|
||||
0.918023532863447,-0.515205295487776,-0.5642053296945,-0.00602353286344737,-0.0239176600611151,-0.560692124132005,-0.590510997925376
|
||||
0.923129550226661,1.03262016372701,1.01116522680603,0.0118704497733385,0.0448548512198665,1.03845846694652,1.09931575827466
|
||||
0.923129550226661,0.381859782258925,0.301449984612702,0.00487044977333873,0.0150633721746665,0.314245969033599,0.332714652297136
|
||||
0.918023532863447,-0.272408599341429,-0.324820914964905,-0.00302353286344714,-0.0132867920220962,-0.328127027921809,-0.345558331766231
|
||||
0.918023532863447,-0.226754579862161,-0.269883188299346,-0.00202353286344714,-0.00888608306610775,-0.273022082268535,-0.287474583737226
|
||||
0.918023532863447,0.903849126847892,0.869867326874846,0.0109764671365528,0.0394710984913298,0.890754692526278,0.938168842232837
|
||||
0.918023532863447,0.162379102254773,0.0659862616502334,0.00197646713655275,0.00424246655240526,0.0955218128304648,0.100606642499834
|
||||
0.918023532863447,-0.223720988594165,-0.26737967038831,-0.00202353286344714,-0.00892954951459952,-0.270683222045707,-0.285014452775087
|
||||
0.923129550226661,0.260108419658216,0.184475101737707,0.00287044977333861,0.00818026681503129,0.197292227508009,0.20884303478398
|
||||
|
@ -0,0 +1,83 @@
|
||||
import numpy as np
|
||||
import os
|
||||
import pandas as pd
|
||||
from statsmodels.tools.testing import Holder
|
||||
|
||||
cur_dir = os.path.dirname(os.path.abspath(__file__))
|
||||
|
||||
results_meth = Holder()
|
||||
results_meth.type = 'ML'
|
||||
results_meth.method = 'BFGS'
|
||||
results_meth.scoring = 3
|
||||
results_meth.start = np.array([
|
||||
1.44771372395646, 0.0615237727637243, 0.604926837329731, 0.98389051740736,
|
||||
6.25859738441389, 0
|
||||
])
|
||||
results_meth.n = 36
|
||||
results_meth.nobs = 36
|
||||
results_meth.df_null = 34
|
||||
results_meth.df_residual = 30
|
||||
results_meth.loglik = 104.148028405343
|
||||
results_meth.vcov = np.array([
|
||||
0.00115682165449043, -0.000665413980696048, -0.000924081767589657,
|
||||
-0.000924126199147583, 0.000941505276523348, -1.44829373972985e-05,
|
||||
-0.000665413980696048, 0.00190019966824938, 4.45163588328844e-06,
|
||||
6.23668249663711e-06, -0.00216418558500309, 4.18754929463506e-05,
|
||||
-0.000924081767589657, 4.45163588328844e-06, 0.0023369966334575,
|
||||
0.000924223263225116, 0.000168988804218447, 1.14762434349836e-07,
|
||||
-0.000924126199147583, 6.23668249663711e-06, 0.000924223263225116,
|
||||
0.00282071714820361, 0.000331499252772628, 1.93773358431975e-07,
|
||||
0.000941505276523348, -0.00216418558500309, 0.000168988804218447,
|
||||
0.000331499252772628, 3.20761137509433, -0.0581708456538647,
|
||||
-1.44829373972985e-05, 4.18754929463506e-05, 1.14762434349836e-07,
|
||||
1.93773358431975e-07, -0.0581708456538647, 0.00107353277853341
|
||||
]).reshape(6, 6, order='F')
|
||||
results_meth.pseudo_r_squared = 0.905194911478503
|
||||
results_meth.y = np.array([
|
||||
0.815, 0.803, 0.803, 0.808, 0.855, 0.813, 0.816, 0.827, 0.829, 0.776,
|
||||
0.786, 0.822, 0.891, 0.894, 0.894, 0.869, 0.914, 0.889, 0.885, 0.898,
|
||||
0.896, 0.86, 0.887, 0.88, 0.936, 0.913, 0.9, 0.912, 0.935, 0.928, 0.915,
|
||||
0.916, 0.929, 0.92, 0.916, 0.926
|
||||
])
|
||||
|
||||
# > cat_items(summ_meth, prefix="results_meth.")
|
||||
# duplicate deleted
|
||||
|
||||
results_meth.residuals_type = 'sweighted2'
|
||||
results_meth.iterations = np.array([
|
||||
12, 3
|
||||
])
|
||||
|
||||
results_meth.table_mean = np.array([
|
||||
1.44224319715775, 0.0698572427112336, 0.607345321898288, 0.973547608125426,
|
||||
0.0340120810079364, 0.0435912797271355, 0.0483424930413969,
|
||||
0.0531104241011462, 42.4038504677562, 1.60255085761448, 12.5633843785881,
|
||||
18.3306314080896, 0, 0.109033850726723, 3.35661710796797e-36,
|
||||
4.71401008973566e-75
|
||||
]).reshape(4, 4, order='F')
|
||||
|
||||
results_meth.table_precision = np.array([
|
||||
8.22828526376512, -0.0347054296138766, 1.79098056245575,
|
||||
0.0327648100640521, 4.59429065633335, -1.05922877459173,
|
||||
4.34223794561173e-06, 0.289495603466561
|
||||
]).reshape(2, 4, order='F')
|
||||
|
||||
results_meth.aic = -196.296056810686
|
||||
results_meth.bic = -186.79494317995
|
||||
|
||||
results_meth.table_mean_oim = np.array([
|
||||
1.44224320770907, 0.069857238768632, 0.607345313356895, 0.973547591731571,
|
||||
0.0340453325782864, 0.0435867955242771, 0.0490089283252544,
|
||||
0.053386889034385, 42.362435567127, 1.60271563734762, 12.3925442590004,
|
||||
18.2357056075048, 0, 0.108997449531221, 2.86797597854623e-35,
|
||||
2.68762966306205e-74
|
||||
]).reshape(4, 4, order='F')
|
||||
|
||||
results_meth.table_precision_oim = np.array([
|
||||
8.22828540005571, -0.0347054322904486, 1.83887205150239,
|
||||
0.0336205378385678, 4.4746372611042, -1.0322688012039,
|
||||
7.65411434417314e-06, 0.301946212204644
|
||||
]).reshape(2, 4, order='F')
|
||||
|
||||
results_meth.resid = pd.read_csv(os.path.join(cur_dir,
|
||||
'resid_methylation.csv'))
|
||||
@ -0,0 +1,399 @@
|
||||
import io
|
||||
import os
|
||||
|
||||
import pytest
|
||||
|
||||
import numpy as np
|
||||
from numpy.testing import assert_allclose, assert_equal
|
||||
import pandas as pd
|
||||
import patsy
|
||||
from statsmodels.api import families
|
||||
from statsmodels.tools.sm_exceptions import (
|
||||
ValueWarning,
|
||||
)
|
||||
from statsmodels.othermod.betareg import BetaModel
|
||||
from .results import results_betareg as resultsb
|
||||
|
||||
links = families.links
|
||||
|
||||
cur_dir = os.path.dirname(os.path.abspath(__file__))
|
||||
res_dir = os.path.join(cur_dir, "results")
|
||||
|
||||
|
||||
# betareg(I(food/income) ~ income + persons, data = FoodExpenditure)
|
||||
_income_estimates_mean = """\
|
||||
varname Estimate StdError zvalue Pr(>|z|)
|
||||
(Intercept) -0.62254806 0.223853539 -2.781051 5.418326e-03
|
||||
income -0.01229884 0.003035585 -4.051556 5.087819e-05
|
||||
persons 0.11846210 0.035340667 3.352005 8.022853e-04"""
|
||||
|
||||
_income_estimates_precision = """\
|
||||
varname Estimate StdError zvalue Pr(>|z|)
|
||||
(phi) 35.60975 8.079598 4.407366 1.046351e-05
|
||||
"""
|
||||
|
||||
_methylation_estimates_mean = """\
|
||||
varname Estimate StdError zvalue Pr(>|z|)
|
||||
(Intercept) 1.44224 0.03401 42.404 2e-16
|
||||
genderM 0.06986 0.04359 1.603 0.109
|
||||
CpGCpG_1 0.60735 0.04834 12.563 2e-16
|
||||
CpGCpG_2 0.97355 0.05311 18.331 2e-16"""
|
||||
|
||||
_methylation_estimates_precision = """\
|
||||
varname Estimate StdError zvalue Pr(>|z|)
|
||||
(Intercept) 8.22829 1.79098 4.594 4.34e-06
|
||||
age -0.03471 0.03276 -1.059 0.289"""
|
||||
|
||||
|
||||
expected_income_mean = pd.read_table(
|
||||
io.StringIO(_income_estimates_mean), sep=r"\s+")
|
||||
expected_income_precision = pd.read_table(
|
||||
io.StringIO(_income_estimates_precision), sep=r"\s+")
|
||||
|
||||
expected_methylation_mean = pd.read_table(
|
||||
io.StringIO(_methylation_estimates_mean), sep=r"\s+")
|
||||
expected_methylation_precision = pd.read_table(
|
||||
io.StringIO(_methylation_estimates_precision), sep=r"\s+")
|
||||
|
||||
income = pd.read_csv(os.path.join(res_dir, 'foodexpenditure.csv'))
|
||||
methylation = pd.read_csv(os.path.join(res_dir, 'methylation-test.csv'))
|
||||
|
||||
|
||||
def check_same(a, b, eps, name):
|
||||
assert np.allclose(a, b, rtol=0.01, atol=eps), \
|
||||
("different from expected", name, list(a), list(b))
|
||||
|
||||
|
||||
def assert_close(a, b, eps):
|
||||
assert np.allclose(a, b, rtol=0.01, atol=eps), (list(a), list(b))
|
||||
|
||||
|
||||
class TestBetaModel:
|
||||
|
||||
@classmethod
|
||||
def setup_class(cls):
|
||||
model = "I(food/income) ~ income + persons"
|
||||
cls.income_fit = BetaModel.from_formula(model, income).fit()
|
||||
|
||||
model = cls.model = "methylation ~ gender + CpG"
|
||||
Z = cls.Z = patsy.dmatrix("~ age", methylation)
|
||||
mod = BetaModel.from_formula(model, methylation, exog_precision=Z,
|
||||
link_precision=links.Identity())
|
||||
cls.meth_fit = mod.fit()
|
||||
mod = BetaModel.from_formula(model, methylation, exog_precision=Z,
|
||||
link_precision=links.Log())
|
||||
cls.meth_log_fit = mod.fit()
|
||||
|
||||
def test_income_coefficients(self):
|
||||
rslt = self.income_fit
|
||||
assert_close(rslt.params[:-1], expected_income_mean['Estimate'], 1e-3)
|
||||
assert_close(rslt.tvalues[:-1], expected_income_mean['zvalue'], 0.1)
|
||||
assert_close(rslt.pvalues[:-1], expected_income_mean['Pr(>|z|)'], 1e-3)
|
||||
|
||||
def test_income_precision(self):
|
||||
|
||||
rslt = self.income_fit
|
||||
# note that we have to exp the phi results for now.
|
||||
assert_close(np.exp(rslt.params[-1:]),
|
||||
expected_income_precision['Estimate'], 1e-3)
|
||||
# yield check_same, rslt.tvalues[-1:],
|
||||
# expected_income_precision['zvalue'], 0.1, "z-score"
|
||||
assert_close(rslt.pvalues[-1:],
|
||||
expected_income_precision['Pr(>|z|)'], 1e-3)
|
||||
|
||||
def test_methylation_coefficients(self):
|
||||
rslt = self.meth_fit
|
||||
assert_close(rslt.params[:-2],
|
||||
expected_methylation_mean['Estimate'], 1e-2)
|
||||
assert_close(rslt.tvalues[:-2],
|
||||
expected_methylation_mean['zvalue'], 0.1)
|
||||
assert_close(rslt.pvalues[:-2],
|
||||
expected_methylation_mean['Pr(>|z|)'], 1e-2)
|
||||
|
||||
def test_methylation_precision(self):
|
||||
# R results are from log link_precision
|
||||
rslt = self.meth_log_fit
|
||||
assert_allclose(rslt.params[-2:],
|
||||
expected_methylation_precision['Estimate'],
|
||||
atol=1e-5, rtol=1e-10)
|
||||
# expected_methylation_precision['Estimate']
|
||||
# yield check_same, links.logit()(rslt.params[-2:]),
|
||||
# expected_methylation_precision['Estimate'], 1e-3, "estimate"
|
||||
# yield check_same, rslt.tvalues[-2:],
|
||||
# expected_methylation_precision['zvalue'], 0.1, "z-score"
|
||||
|
||||
def test_precision_formula(self):
|
||||
m = BetaModel.from_formula(self.model, methylation,
|
||||
exog_precision_formula='~ age',
|
||||
link_precision=links.Identity())
|
||||
rslt = m.fit()
|
||||
assert_close(rslt.params, self.meth_fit.params, 1e-10)
|
||||
assert isinstance(rslt.params, pd.Series)
|
||||
|
||||
with pytest.warns(ValueWarning, match="unknown kwargs"):
|
||||
BetaModel.from_formula(self.model, methylation,
|
||||
exog_precision_formula='~ age',
|
||||
link_precision=links.Identity(),
|
||||
junk=False)
|
||||
|
||||
def test_scores(self):
|
||||
model, Z = self.model, self.Z
|
||||
for link in (links.Identity(), links.Log()):
|
||||
mod2 = BetaModel.from_formula(model, methylation, exog_precision=Z,
|
||||
link_precision=link)
|
||||
rslt_m = mod2.fit()
|
||||
|
||||
# evaluate away from optimum to get larger score
|
||||
analytical = rslt_m.model.score(rslt_m.params * 1.01)
|
||||
numerical = rslt_m.model._score_check(rslt_m.params * 1.01)
|
||||
assert_allclose(analytical, numerical, rtol=1e-6, atol=1e-6)
|
||||
assert_allclose(link.inverse(analytical[3:]),
|
||||
link.inverse(numerical[3:]), rtol=5e-7, atol=5e-6)
|
||||
|
||||
def test_results_other(self):
|
||||
|
||||
rslt = self.meth_fit
|
||||
distr = rslt.get_distribution()
|
||||
mean, var = distr.stats()
|
||||
assert_allclose(rslt.fittedvalues, mean, rtol=1e-13)
|
||||
assert_allclose(rslt.model._predict_var(rslt.params), var, rtol=1e-13)
|
||||
resid = rslt.model.endog - mean
|
||||
assert_allclose(rslt.resid, resid, rtol=1e-12)
|
||||
assert_allclose(rslt.resid_pearson, resid / np.sqrt(var), rtol=1e-12)
|
||||
|
||||
|
||||
class TestBetaMeth():
|
||||
|
||||
@classmethod
|
||||
def setup_class(cls):
|
||||
formula = "methylation ~ gender + CpG"
|
||||
mod = BetaModel.from_formula(formula, methylation,
|
||||
exog_precision_formula="~ age",
|
||||
link_precision=links.Log())
|
||||
cls.res1 = mod.fit(cov_type="eim")
|
||||
cls.res2 = resultsb.results_meth
|
||||
|
||||
def test_basic(self):
|
||||
res1 = self.res1
|
||||
res2 = self.res2
|
||||
|
||||
k_mean = 4
|
||||
p, se, zv, pv = res2.table_mean.T
|
||||
assert_allclose(res1.params[:k_mean], p, rtol=1e-6)
|
||||
assert_allclose(res1.bse[:k_mean], se, rtol=1e-6)
|
||||
assert_allclose(res1.tvalues[:k_mean], zv, rtol=1e-6)
|
||||
assert_allclose(res1.pvalues[:k_mean], pv, rtol=1e-5)
|
||||
|
||||
p, se, zv, pv = res2.table_precision.T
|
||||
assert_allclose(res1.params[k_mean:], p, rtol=1e-6)
|
||||
assert_allclose(res1.bse[k_mean:], se, rtol=1e-6)
|
||||
assert_allclose(res1.tvalues[k_mean:], zv, rtol=1e-6)
|
||||
assert_allclose(res1.pvalues[k_mean:], pv, rtol=1e-5)
|
||||
|
||||
assert_allclose(res1.llf, res2.loglik, rtol=1e-10)
|
||||
assert_allclose(res1.aic, res2.aic, rtol=1e-10)
|
||||
assert_allclose(res1.bic, res2.bic, rtol=1e-10)
|
||||
# dofferent definitions for prsquared
|
||||
assert_allclose(res1.prsquared, res2.pseudo_r_squared, atol=0.01)
|
||||
assert_equal(res1.df_resid, res2.df_residual)
|
||||
assert_equal(res1.nobs, res2.nobs)
|
||||
|
||||
# null model compared to R betareg and lmtest
|
||||
df_c = res1.df_resid_null - res1.df_resid
|
||||
assert_equal(res1.k_null, 2)
|
||||
|
||||
# > lrt = lrtest(res_meth_null, res_meth) # results from R
|
||||
pv = 7.21872953868659e-18
|
||||
lln = 60.88809589492269
|
||||
llf = 104.14802840534323
|
||||
chisq = 86.51986502084107
|
||||
dfc = 4
|
||||
# stats.chi2.sf(86.51986502093865, 4)
|
||||
assert_equal(df_c, dfc)
|
||||
assert_allclose(res1.llf, llf, rtol=1e-10)
|
||||
assert_allclose(res1.llnull, lln, rtol=1e-10)
|
||||
assert_allclose(res1.llr, chisq, rtol=1e-10)
|
||||
assert_allclose(res1.llr_pvalue, pv, rtol=1e-6)
|
||||
|
||||
def test_resid(self):
|
||||
res1 = self.res1
|
||||
res2 = self.res2
|
||||
assert_allclose(res1.fittedvalues, res2.resid['fittedvalues'],
|
||||
rtol=1e-8)
|
||||
assert_allclose(res1.resid, res2.resid['response'],
|
||||
atol=1e-8, rtol=1e-8)
|
||||
|
||||
def test_oim(self):
|
||||
# estimate with default oim, cov_type nonrobust
|
||||
res1 = self.res1.model.fit()
|
||||
res2 = self.res2
|
||||
|
||||
k_mean = 4
|
||||
# R betareg uses numerical derivatives from bfgs, has lower precision
|
||||
p, se, zv, pv = res2.table_mean_oim.T
|
||||
assert_allclose(res1.params[:k_mean], p, rtol=1e-6)
|
||||
assert_allclose(res1.bse[:k_mean], se, rtol=1e-5)
|
||||
assert_allclose(res1.tvalues[:k_mean], zv, rtol=1e-5)
|
||||
assert_allclose(res1.pvalues[:k_mean], pv, atol=1e-6, rtol=1e-5)
|
||||
|
||||
p, se, zv, pv = res2.table_precision_oim.T
|
||||
assert_allclose(res1.params[k_mean:], p, rtol=1e-6)
|
||||
assert_allclose(res1.bse[k_mean:], se, rtol=1e-3)
|
||||
assert_allclose(res1.tvalues[k_mean:], zv, rtol=1e-3)
|
||||
assert_allclose(res1.pvalues[k_mean:], pv, rtol=1e-2)
|
||||
|
||||
def test_predict_distribution(self):
|
||||
res1 = self.res1
|
||||
mean = res1.predict()
|
||||
var_ = res1.model._predict_var(res1.params)
|
||||
distr = res1.get_distribution()
|
||||
m2, v2 = distr.stats()
|
||||
assert_allclose(mean, m2, rtol=1e-13)
|
||||
assert_allclose(var_, v2, rtol=1e-13)
|
||||
|
||||
# from R: > predict(res_meth, type="variance")
|
||||
var_r6 = [
|
||||
3.1090848852102e-04, 2.4509604000073e-04, 3.7199753140565e-04,
|
||||
2.8088261358738e-04, 2.7561111800350e-04, 3.3929220526847e-04]
|
||||
n = 6
|
||||
assert_allclose(v2[:n], var_r6, rtol=1e-7)
|
||||
|
||||
ex = res1.model.exog[:n]
|
||||
ex_prec = res1.model.exog_precision[:n]
|
||||
mean6 = res1.predict(ex, transform=False)
|
||||
prec = res1.predict(which="precision")
|
||||
# todo: prec6 wrong exog if not used as keyword, no exception raised
|
||||
prec6 = res1.predict(exog_precision=ex_prec, which="precision",
|
||||
transform=False)
|
||||
var6 = res1.model._predict_var(res1.params, exog=ex,
|
||||
exog_precision=ex_prec)
|
||||
|
||||
assert_allclose(mean6, mean[:n], rtol=1e-13)
|
||||
assert_allclose(prec6, prec[:n], rtol=1e-13)
|
||||
assert_allclose(var6, var_[:n], rtol=1e-13)
|
||||
assert_allclose(var6, var_r6, rtol=1e-7)
|
||||
|
||||
distr6 = res1.model.get_distribution(res1.params,
|
||||
exog=ex, exog_precision=ex_prec)
|
||||
m26, v26 = distr6.stats()
|
||||
assert_allclose(m26, m2[:n], rtol=1e-13)
|
||||
assert_allclose(v26, v2[:n], rtol=1e-13)
|
||||
|
||||
distr6f = res1.get_distribution(exog=ex, exog_precision=ex_prec,
|
||||
transform=False)
|
||||
m26, v26 = distr6f.stats()
|
||||
assert_allclose(m26, m2[:n], rtol=1e-13)
|
||||
assert_allclose(v26, v2[:n], rtol=1e-13)
|
||||
|
||||
# check formula transform works for predict, currently mean only
|
||||
df6 = methylation.iloc[:6]
|
||||
mean6f = res1.predict(df6)
|
||||
# todo: prec6 wrong exog if not used as keyword, no exception raised
|
||||
# formula not supported for exog_precision in predict
|
||||
# prec6f = res1.predict(exog_precision=ex_prec, which="precision")
|
||||
assert_allclose(mean6f, mean[:n], rtol=1e-13)
|
||||
# assert_allclose(prec6f, prec[:n], rtol=1e-13)
|
||||
|
||||
distr6f = res1.get_distribution(exog=df6, exog_precision=ex_prec)
|
||||
m26, v26 = distr6f.stats()
|
||||
assert_allclose(m26, m2[:n], rtol=1e-13)
|
||||
assert_allclose(v26, v2[:n], rtol=1e-13)
|
||||
# check that we don't have pandas in distr
|
||||
assert isinstance(distr6f.args[0], np.ndarray)
|
||||
|
||||
# minimal checks for get_prediction
|
||||
pma = res1.get_prediction(which="mean", average=True)
|
||||
dfma = pma.summary_frame()
|
||||
assert_allclose(pma.predicted, mean.mean(), rtol=1e-13)
|
||||
assert_equal(dfma.shape, (1, 4))
|
||||
pm = res1.get_prediction(exog=df6, which="mean", average=False)
|
||||
dfm = pm.summary_frame()
|
||||
assert_allclose(pm.predicted, mean6, rtol=1e-13)
|
||||
assert_equal(dfm.shape, (6, 4))
|
||||
pv = res1.get_prediction(exog=df6, exog_precision=ex_prec,
|
||||
which="var", average=False)
|
||||
dfv = pv.summary_frame()
|
||||
assert_allclose(pv.predicted, var6, rtol=1e-13)
|
||||
assert_equal(dfv.shape, (6, 4))
|
||||
# smoke tests
|
||||
res1.get_prediction(which="linear", average=False)
|
||||
res1.get_prediction(which="precision", average=True)
|
||||
res1.get_prediction(exog_precision=ex_prec, which="precision",
|
||||
average=False)
|
||||
res1.get_prediction(which="linear-precision", average=True)
|
||||
|
||||
# test agg_weights
|
||||
pm = res1.get_prediction(exog=df6, which="mean", average=True)
|
||||
dfm = pm.summary_frame()
|
||||
aw = np.zeros(len(res1.model.endog))
|
||||
aw[:6] = 1
|
||||
aw /= aw.mean()
|
||||
pm6 = res1.get_prediction(exog=df6, which="mean", average=True)
|
||||
dfm6 = pm6.summary_frame()
|
||||
pmw = res1.get_prediction(which="mean", average=True, agg_weights=aw)
|
||||
dfmw = pmw.summary_frame()
|
||||
assert_allclose(pmw.predicted, pm6.predicted, rtol=1e-13)
|
||||
assert_allclose(dfmw, dfm6, rtol=1e-13)
|
||||
|
||||
|
||||
class TestBetaIncome():
|
||||
|
||||
@classmethod
|
||||
def setup_class(cls):
|
||||
|
||||
formula = "I(food/income) ~ income + persons"
|
||||
exog_prec = patsy.dmatrix("~ persons", income)
|
||||
mod_income = BetaModel.from_formula(
|
||||
formula,
|
||||
income,
|
||||
exog_precision=exog_prec,
|
||||
link_precision=links.Log()
|
||||
)
|
||||
res_income = mod_income.fit(method="newton")
|
||||
|
||||
mod_restricted = BetaModel.from_formula(
|
||||
formula,
|
||||
income,
|
||||
link_precision=links.Log()
|
||||
)
|
||||
res_restricted = mod_restricted.fit(method="newton")
|
||||
|
||||
cls.res1 = res_income
|
||||
cls.resr = res_restricted
|
||||
|
||||
def test_score_test(self):
|
||||
res1 = self.res1
|
||||
resr = self.resr
|
||||
params_restr = np.concatenate([resr.params, [0]])
|
||||
r_matrix = np.zeros((1, len(params_restr)))
|
||||
r_matrix[0, -1] = 1
|
||||
exog_prec_extra = res1.model.exog_precision[:, 1:]
|
||||
|
||||
from statsmodels.base._parameter_inference import score_test
|
||||
sc1 = score_test(res1, params_constrained=params_restr,
|
||||
k_constraints=1)
|
||||
sc2 = score_test(resr, exog_extra=(None, exog_prec_extra))
|
||||
assert_allclose(sc2[:2], sc1[:2])
|
||||
|
||||
sc1_hc = score_test(res1, params_constrained=params_restr,
|
||||
k_constraints=1, r_matrix=r_matrix, cov_type="HC0")
|
||||
sc2_hc = score_test(resr, exog_extra=(None, exog_prec_extra),
|
||||
cov_type="HC0")
|
||||
assert_allclose(sc2_hc[:2], sc1_hc[:2])
|
||||
|
||||
def test_influence(self):
|
||||
# currently only smoke test
|
||||
res1 = self.res1
|
||||
from statsmodels.stats.outliers_influence import MLEInfluence
|
||||
|
||||
influ0 = MLEInfluence(res1)
|
||||
influ = res1.get_influence()
|
||||
attrs = ['cooks_distance', 'd_fittedvalues', 'd_fittedvalues_scaled',
|
||||
'd_params', 'dfbetas', 'hat_matrix_diag', 'resid_studentized'
|
||||
]
|
||||
for attr in attrs:
|
||||
getattr(influ, attr)
|
||||
|
||||
frame = influ.summary_frame()
|
||||
frame0 = influ0.summary_frame()
|
||||
assert_allclose(frame, frame0, rtol=1e-13, atol=1e-13)
|
||||
Reference in New Issue
Block a user