菜鸟求助~~~急！！！

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

我的计量基础很差，最近做毕业论文时遇到了很大的问题，想请各位高手予以指点：

我是要根据一组面板数据做一个probit或者logit回归，首先对这些指标进行过corr，相关系数大多都在0.1以内，少数几个在0.5以下，做VIF分析，也都在2一下，是不是可以不考虑共线性了呢？

但是回归的结果感觉不大好，大多数系数的z值都很小，而且其它相关指标我的理解也很模糊，这两天虽然一直在网上查阅相关资料，但总还是看得云里雾里的。先贴一个试验结果发上来：

Fitting comparison model:

Iteration 0:   log likelihood = -198.93324
Iteration 1:   log likelihood =   -89.5573
Iteration 2:   log likelihood = -71.344916
Iteration 3:   log likelihood = -67.369216
Iteration 4:   log likelihood = -66.924545
Iteration 5:   log likelihood = -66.917215
Iteration 6:   log likelihood = -66.917213

Fitting full model:

rho =  0.0     log likelihood = -92.512473
rho =  0.1     log likelihood = -55.467123
rho =  0.2     log likelihood = -41.736322
rho =  0.3     log likelihood = -33.637198
rho =  0.4     log likelihood = -27.967127
rho =  0.5     log likelihood = -23.714651
rho =  0.6     log likelihood = -20.197696
rho =  0.7     log likelihood = -17.289147
rho =  0.8     log likelihood = -14.803667

Iteration 0:   log likelihood = -16.824675 
Iteration 1:   log likelihood = -13.217479 
Iteration 2:   log likelihood = -12.605991 
Iteration 3:   log likelihood = -12.587627 
Iteration 4:   log likelihood = -12.587611 
Iteration 5:   log likelihood = -12.587611 

Random-effects probit regression                Number of obs      =       287
Group variable (i): country                     Number of groups   =        13

Random effects u_i ~ Gaussian                   Obs per group: min =        15
                                                               avg =      22.1
                                                               max =        24

                                                Wald chi2(11)      =      7.91
Log likelihood  = -12.587611                    Prob > chi2        =    0.7214

------------------------------------------------------------------------------
           y |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
         rer |  -3.475526   3.158579    -1.10   0.271    -9.666228    2.715176
       res_g |    .825807   1.651825     0.50   0.617    -2.411711    4.063325
      m2_res |   .9094768   2.506475     0.36   0.717    -4.003125    5.822078
    m2_res_g |   1.246309   1.961459     0.64   0.525     -2.59808    5.090698
       gdp_g |   1.384571   1.648769     0.84   0.401    -1.846958    4.616099
      dc_gdp |   3.991885   2.858283     1.40   0.163    -1.610247    9.594017
      ca_gdp |  -7.376085   3.668452    -2.01   0.044    -14.56612   -.1860506
       rl_rd |  -1.043828   2.217364    -0.47   0.638    -5.389782    3.302125
       exp_g |  -.5325338   1.768384    -0.30   0.763    -3.998502    2.933435
        im_g |   .6575589   1.705087     0.39   0.700    -2.684351    3.999469
         inf |   .2479812    1.57941     0.16   0.875    -2.847606    3.343568
-------------+----------------------------------------------------------------
    /lnsig2u |   2.113526   .4843254                      1.164265    3.062786
-------------+----------------------------------------------------------------
     sigma_u |   2.877043   .6967125                      1.789852    4.624615
         rho |   .8922109   .0465779                      .7621069    .9553313
------------------------------------------------------------------------------
Likelihood-ratio test of rho=0: chibar2(01) =   108.66 Prob >= chibar2 = 0.000

想请各位给我指点一下：     1、 Log likelihood  = -12.587611         Wald chi2(11)      Prob > chi2        =    0.7214

                                                     /lnsig2u      sigma_u          rho |  

                                                        Likelihood-ratio test of rho=0: chibar2(01) =   108.66 Prob >= chibar2 = 0.000

                                                      这些检验都应该怎么去看？

                                          2、像这种表现不好的模型可以从哪些方面改进呢？或者说其问题可能出在哪里了呢？

                 快答辩了，可是这个实证困扰我很久了，真头大~~

                  期盼您的解答，不胜感激！！！

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0 回帖

关键词：Likelihood regression Comparison iteration regressio 求助 菜鸟

个人感觉理论是关键……数学和统计只是一种工具罢了……构建模型都是应该在理论上面的而不是自己的凭空觉得手里有哪些就放进去都试试吧……

我是根据实际理论和相关文献确定的方法，可是效果总不好，不得要领~~