关于boxcox检验的问题

是 否 +2 论坛币

k人 参与回答

经管之家送您一份

应届毕业生专属福利!

求职就业群

赵安豆老师微信：zhaoandou666

经管之家联合CDA

送您一个全额奖学金名额~ !

立即领取

感谢您参与论坛问题回答

经管之家送您两个论坛币！

+2 论坛币

. boxcox rmedicalexpenditureper rgorvenmentrevenueper rtransferper, notrans( above
> 65proportion below15proportion morbilityper1000 mortalityper1000  nonprofitbedsp
> er10000 urbanproportion femaleproportion juniorcollegeedupropotion unmarried ins
> urancedcoverage  year2003 year2004 year2005 year2006) model(lambda)
Fitting comparison model

Iteration 0:   log likelihood = -899.82201 
Iteration 1:   log likelihood =  -808.6929 
Iteration 2:   log likelihood =  -808.2444 
Iteration 3:   log likelihood = -808.24417 
Iteration 4:   log likelihood = -808.24417 

Fitting full model

Iteration 0:   log likelihood =  -712.2262 
Iteration 1:   log likelihood = -644.48226 
Iteration 2:   log likelihood = -640.36162 
Iteration 3:   log likelihood = -623.09097 
Iteration 4:   log likelihood = -622.92809 
Iteration 5:   log likelihood = -622.92764 
Iteration 6:   log likelihood = -622.92764 

                                                  Number of obs   =        155
                                                  LR chi2(16)     =     370.63
Log likelihood = -622.92764                       Prob > chi2     =      0.000
 
------------------------------------------------------------------------------
rmedicalex~r |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
     /lambda |   .0405063    .073433     0.55   0.581    -.1034199    .1844324
------------------------------------------------------------------------------
 
Estimates of scale-variant parameters
----------------------------
             |      Coef.
-------------+--------------
Notrans      |
above65pro~n |  -5.976282
below15pro~n |    .485161
morbili~1000 |   .0273048
mortali~1000 |   10.51786
nonpro~10000 |  -.0041632
urbanpropo~n |  -1.656319
femaleprop~n |   10.72626
juniorcoll~n |   3.717347
   unmarried |   1.858071
insuranced~e |  -.8560323
    year2003 |    .053958
    year2004 |  -.0405212
    year2005 |   .0450074
    year2006 |   .0511788
       _cons |  -9.382694
-------------+--------------
Trans        |
rgorvenmen~r |   .7286858
rtransferper |   .4800122
-------------+--------------
      /sigma |   .2112387
----------------------------
 
---------------------------------------------------------
   Test         Restricted     LR statistic      P-value
    H0:       log likelihood       chi2       Prob > chi2
---------------------------------------------------------
lambda = -1     -677.88107       109.91           0.000
lambda =  0     -623.07519         0.30           0.587
lambda =  1      -712.2262       178.60           0.000
---------------------------------------------------------
上面是我做的boxcox检验。

检验的目的是，看变量采用log的形式，还是直接数值的形式。

请问高手，这个结果是不是代表我两边取log，是不能被拒绝。因此采用两边用log形式更好？

非常感谢！

[此贴子已经被蓝色于2009-5-12 15:11:54编辑过]

扫码加我 拉你入群

请注明：姓名-公司-职位

以便审核进群资格，未注明则拒绝

分享0 收藏1 回帖

关键词：boxcox box Cox expenditure Likelihood 检验 boxcox

当样本量较大的时候样本渐进不从正态分布的，所以应该不需要做box-cox变换