R: BMSパッケージ

R-bloggersの記事BMS 0.3.0 ReleasedにあったBMSをためしてみた。Bayesian Model Averagingのパッケージである。

Tutorialにあった例をためしてみる。

> library(BMS)
> # dataflsを使用。
> data(datafls)
> # 最初の列に目的変数、それ以外の列に説明変数がはいっている。
> head(datafls)
          y Abslat Spanish French Brit WarDummy LatAmerica SubSahara OutwarOr Area PrScEnroll
DZ 0.013690 36.720       0      1    0        0          0         0        1 2382       0.46
AR 0.006421 36.676       1      0    0        1          1         0        0 2767       0.98
AU 0.018897 32.219       0      0    1        0          0         0        0 7687       1.00
AT 0.029221 48.231       0      0    0        0          0         0        0   84       1.00
BE 0.028571 50.837       0      0    0        0          0         0        1   31       1.00
BO 0.012652 15.190       1      0    0        0          1         0        0 1099       0.64
   LifeExp  GDP60 Mining EcoOrg YrsOpen Age Buddha Catholic Confucian EthnoL Hindu Jewish
DZ    47.3 7.4390  0.196      0   0.000  84      0     0.00         0   0.43     0  0.005
AR    65.2 8.0999  0.023      5   0.089   0      0     0.90         0   0.31     0  0.020
AU    70.7 8.9721  0.038      5   0.689   0      0     0.26         0   0.32     0  0.000
AT    68.8 8.5471  0.004      4   0.778   0      0     0.85         0   0.13     0  0.000
BE    69.7 8.6223  0.000      5   0.778  88      0     0.75         0   0.55     0  0.000
BO    42.8 7.0058  0.075      3   0.733   0      0     0.95         0   0.68     0  0.000
   Muslim PrExports Protestants RuleofLaw     Popg  WorkPop LabForce HighEnroll PublEdupct
DZ   0.99     0.933       0.005   0.33333 0.028417 -1.33030 2855.520      0.003     0.0297
AR   0.00     0.861       0.020   0.33333 0.015255 -0.93293 8110.334      0.109     0.0223
AU   0.00     0.811       0.500   1.00000 0.016453 -0.90190 4185.827      0.131     0.0217
AT   0.00     0.198       0.060   1.00000 0.002369 -0.73355 3384.449      0.080     0.0248
BE   0.00     0.238       0.130   1.00000 0.002494 -0.95399 3525.736      0.090     0.0463
BO   0.00     0.969       0.030   0.16667 0.025138 -1.06310 1184.031      0.036     0.0158
   RevnCoup PolRights CivlLib English Foreign RFEXDist EquipInv NequipInv   stdBMP BlMktPm
DZ    0.123    5.8333  5.8889    0.00   0.836      190  0.05070   0.19070 186.5555   0.131
AR    0.960    3.9444  3.5556    0.00   0.836      113  0.01600   0.12000  66.8497   0.160
AU    0.000    1.0000  1.0000    0.95   0.000      129  0.09500   0.22100   0.0010   0.000
AT    0.000    1.0000  1.0000    0.00   0.980      100  0.10059   0.13391   0.0010   0.000
BE    0.000    1.0000  1.0000    0.00   0.345      100  0.06348   0.17082   0.0010   0.000
BO    1.190    4.1111  3.6667    0.00   0.372      181  0.01400   0.11950 236.0656   0.030
>
> # bms()を実行
> mfls = bms(datafls, burn=100000, iter=200000, g="BRIC", mprior="uniform", nmodel=2000, mcmc="bd", user.int=FALSE)
> coef(mfls,exact=TRUE)
                    PIP     Post Mean      Post SD Cond.Pos.Sign Idx
GDP60       1.000000000 -1.621129e-02 2.942707e-03    0.00000000  12
Confucian   0.999685123  5.621758e-02 1.251057e-02    1.00000000  19
EquipInv    0.964621576  1.661892e-01 5.935999e-02    1.00000000  38
LifeExp     0.960776010  8.416336e-04 3.063505e-04    1.00000000  11
SubSahara   0.785279889 -1.199071e-02 7.753254e-03    0.00000000   7
Muslim      0.683064734  8.750278e-03 7.050511e-03    1.00000000  23
RuleofLaw   0.554828881  8.562602e-03 8.506714e-03    1.00000000  26
EcoOrg      0.499971849  1.371179e-03 1.502426e-03    1.00000000  14
YrsOpen     0.486797992  7.130821e-03 8.058750e-03    1.00000000  15
Protestants 0.471505006 -5.918911e-03 7.052854e-03    0.00000000  25
NequipInv   0.446358378  2.603754e-02 3.210288e-02    1.00000000  39
Mining      0.401525239  1.611649e-02 2.199300e-02    1.00000000  13
PrScEnroll  0.152114791  3.239182e-03 8.461640e-03    0.99824948  10
LatAmerica  0.150278800 -1.304647e-03 3.425278e-03    0.01371394   6
Buddha      0.130511953  1.648206e-03 4.764036e-03    1.00000000  17
BlMktPm     0.119847685 -9.287419e-04 2.803381e-03    0.00000000  41
Catholic    0.093368040 -3.963352e-04 2.647377e-03    0.23995138  18
Hindu       0.072448224 -1.715739e-03 7.018795e-03    0.01916108  21
CivlLib     0.066299920 -1.555296e-04 6.606950e-04    0.00000000  34
PrExports   0.044672181 -4.587887e-04 2.429760e-03    0.00000000  24
PolRights   0.040367741 -6.530202e-05 3.703630e-04    0.00000000  33
RFEXDist    0.039546498 -2.356242e-06 1.329685e-05    0.00921559  37
Age         0.028332106 -1.269423e-06 8.870516e-06    0.00000000  16
WarDummy    0.026950244 -1.056351e-04 7.549343e-04    0.00000000   5
Foreign     0.025532677  1.315906e-04 9.637474e-04    0.95633624  36
English     0.024039295 -1.541018e-04 1.178185e-03    0.00000000  35
LabForce    0.020872752  1.275105e-09 1.347657e-08    0.81071150  29
EthnoL      0.015872042  8.662374e-05 8.790468e-04    0.96602763  20
Spanish     0.014723100  5.949540e-05 7.618993e-04    0.87910071   2
stdBMP      0.014285817 -2.146859e-07 2.289706e-06    0.00000000  40
French      0.013409248  5.950078e-05 6.494679e-04    1.00000000   3
Popg        0.009484490  1.635049e-03 2.322621e-02    0.98568876  27
Abslat      0.009098257 -7.654455e-09 1.425345e-05    0.37687900   1
WorkPop     0.008734579 -5.551015e-05 9.364190e-04    0.10010695  28
HighEnroll  0.007760023 -1.860574e-04 3.178367e-03    0.00000000  30
Jewish      0.006978195 -6.079308e-05 1.250914e-03    0.13881329  22
RevnCoup    0.006729709  5.182697e-06 4.985380e-04    0.61177202  32
Brit        0.006566336 -1.358732e-05 2.494794e-04    0.00000000   4
OutwarOr    0.005967004 -1.027491e-05 2.248706e-04    0.08938727   8
Area        0.005412366 -1.522509e-09 4.108108e-08    0.11342615   9
PublEdupct  0.004680216  6.653055e-05 9.143782e-03    0.52782801  31

もっとくわしいTutorial: Bayesian Model Averaging with BMS

タグ：R