模型操作_动态Probit模型及Stata操作应用

我们用R包cquad来说明Bartolucci和Nigro(2010)提出的用于分析条件极大似然估计的面板二元选择模型。该包还允许我们估计动态面板logit模型。我们还演示了Stata中可用的这个包的简化版本。这个包的主要功能的使用是基于使用劳动力市场数据的例子。

1 简介随着可获得的面板数据集的数量不断增加以及相关统计和计量模型的最新发展，用于估计二元面板数据的非线性模型的现成软件现在在应用研究中是必不可少的。在本文中，我们举例说明了cquad，可以在http://cran.r-project.org/web/packages/cquad/index.html获得，它是一个用于固定效应二进制面板数据模型CML估计的综合R包。 特别是cquad包含了估算静态logit模型(Chamberlain, 1980)和Bartolucci等人(2015)最近提出的动态量化宽松模型的函数。 Stata还可以通过https://ideas.repec.org/c/boc/bocode/s457891.html获得R包cquad的一个版本，包括它的主要功能。由于cquad为二元因变量实现了非线性面板数据模型的固定效应估计，因此它补充了面板数据计量经济学中现有的R包。最重要的是，它与plm包密切相关， plm包提供了广泛的函数集，用于估计线性面板数据模型，包括静态和动态估计。此外，cquad与plm在数据组结构、模型的公式等方面具有相同的特点。下面将描述用于R的包cquad和用于Stata的相应模块的主要功能。为了描述cquad的功能，我们使用了从美国全国青年纵向调查中提取的工会工人数据。为了演示R包，我们使用了与Wooldridge(2005)中相同的数据，这些数据可以从http://qed.econ.queensu.ca/jae/datasets/wooldridge001/获得， 而对于Stata模块，我们使用Stata存储库中已经可用的类似数据。 2 Stata操作应用 在下面，我们通过一个基于工会工人数据的示例来说明Stata模块cquad，该示例数据经常用于说明动态面板Probit数据模型(Stewart, 2006;Lucchetti和Pigini, 2015)。 下面加载数据集，然后描述已经该面板结构，并列出数据集中出现的变量 。 代码为：

use "E:\stata\data\union.dta"(NLS Women 14-24 in 1968)edxtdes

结果为：

. clear. use "E:\stata\data\union.dta"(NLS Women 14-24 in 1968). ed.  xtdes  idcode:  1, 2, ..., 5159                                   n =       4434    year:  70, 71, ..., 88                                   T =         12           Delta(year) = 1 unit           Span(year)  = 19 periods           (idcode*year uniquely identifies each observation)Distribution of T_i:   min      5%     25%       50%       75%     95%     max                         1       1       3         6         8      11      12     Freq.  Percent    Cum. |  Pattern ---------------------------+---------------------      190      4.29    4.29 |  1111...11.1.11.1.11      129      2.91    7.19 |  .......11.1.11.1.11       93      2.10    9.29 |  1..................       78      1.76   11.05 |  .......1...........       68      1.53   12.58 |  ..11...11.1.11.1.11       64      1.44   14.03 |  ...1...11.1.11.1.11       60      1.35   15.38 |  .111...11.1.11.1.11       52      1.17   16.55 |  11.................       52      1.17   17.73 |  1111...............     3648     82.27  100.00 | (other patterns) ---------------------------+---------------------     4434    100.00         |  XXXX...XX.X.XX.X.XX. descrContains data from E:\stata\data\union.dta  obs:        26,200                          NLS Women 14-24 in 1968 vars:            10                          27 Oct 2006 13:51 size:       288,200                          -------------------------------------------------------------------------------------------------------------------------              storage   display    valuevariable name   type    format     label      variable label-------------------------------------------------------------------------------------------------------------------------idcode          int     %8.0g                 NLS IDyear            byte    %8.0g                 interview yearage             byte    %8.0g                 age in current yeargrade           byte    %8.0g                 current grade completednot_smsa        byte    %8.0g                 1 if not SMSAsouth           byte    %8.0g                 1 if southunion           byte    %8.0g                 1 if uniont0              byte    %9.0g                 southXt         byte    %9.0g                 black           byte    %8.0g                 race black-------------------------------------------------------------------------------------------------------------------------Sorted by: idcode  year

该数据集由4434名女性组成，输出的最后一部分报告了变量的描述，其中union将是我们练习中的因变量，而不是smsa，教育年限和南方的协变量集，而black将被排除在分析之外，因为它的时不变的性质。我们首先演示命令cquadbasic，其中我们使用xi和i.year声明在模型规范中包含时间虚拟变量。命令为：

. xi: cquadbasic union idcode age grade south not_smsa i.year

结果为：

. xi: cquadbasic union idcode age grade south not_smsa i.yeari.year            _Iyear_70-88        (naturally coded; _Iyear_70 omitted)Fit (simplified) quadratic exponential model by Conditional Maximum Likelihoodsee Bartolucci & Nigro (2010), Econometrica      |      lk       lk-lk0------+--------------------------------------------------    1 | -3439.9096  1.000e+10    2 | -3071.6412  368.26839    3 | -3069.0539  2.5872579    4 | -3069.0534  .00050444    5 | -3069.0534  5.684e-11             |      est.      s.e.      t-stat.    p-value-------------+--------------------------------------------------         age |  .17670917   .1192216  1.4821908  .13828953       grade | -.03658997  .04586492 -.79777692  .42499996       south |  -.5191613  .13732314 -3.7805814  .00015646    not_smsa |  .12631127  .13146408   .9608044  .33665052   _Iyear_71 |  1.5208636   1.035464  1.4687749  .14189386   _Iyear_72 |  1.1096837  .91812295  1.2086439  .22679967   _Iyear_73 |  .90256541  .79733234  1.1319814  .25764224   _Iyear_77 |   .1829554   .3308496  .55298662  .58027258   _Iyear_78 |  .17904624  .21288676   .8410398  .40032564   _Iyear_80 |  .46950024   .0846961  5.5433514  2.967e-08   _Iyear_82 | -.40988205  .28664089 -1.4299497  .15273146   _Iyear_83 | -.95438994  .40301052 -2.3681514  .01787722   _Iyear_85 | -.73765258  .63762525 -1.1568748  .24732352   _Iyear_87 | -1.3247366  .87641421 -1.5115416  .13065051   _Iyear_88 | -.93795347  1.0381108 -.90351959  .36625019       y-lag |  1.5332567  .06307817  24.307248          0

上述结果首先报告了迭代日志，然后以标准方式显示估计输出，其中第一列报告基本模型的估计系数，以及渐近标准误差、相关的t统计数据和p值模型的扩展版本也可以适用于类似的方式，使用命令

cquadext union idcode age grade south not_smsa _Iyear_72 _Iyear_73_Iyear_77 _Iyear_78 _Iyear_80 _Iyear_82 _Iyear_83 _Iyear_85 _Iyear_87

结果为：

. cquadext union idcode age grade south not_smsa _Iyear_72 _Iyear_73 _Iyear_77 _Iyear_78 _Iyear_80 _Iyear_82 _Iyear_83 _I> year_85 _Iyear_87Fit quadratic exponential model by Conditional Maximum Likelihoodsee Bartolucci & Nigro (2010), Econometrica      |      lk       lk-lk0------+--------------------------------------------------    1 | -3439.9096  1.000e+10    2 | -3062.7193  377.19031    3 | -3059.9982  2.7210475    4 | -3059.9975  .00070056    5 | -3059.9975  1.597e-09               |      est.      s.e.      t-stat.    p-value---------------+--------------------------------------------------           age |  .17308473  .11933765  1.4503782  .14695309         grade | -.04047509   .0465145 -.87016079  .38421254         south | -.51184847  .13953697 -3.6681926  .00024427      not_smsa |  .17524652  .13523937  1.2958248  .19503586     _Iyear_72 |  -.4644361   .1964388 -2.3642789  .01806521     _Iyear_73 | -.65950516  .27895047 -2.3642375  .01806723     _Iyear_77 | -1.3784265  .72358421 -1.9049981  .05678032     _Iyear_78 | -1.3701126  .84614133 -1.6192479  .10539394     _Iyear_80 | -1.1167485  1.0780889 -1.0358595  .30026771     _Iyear_82 | -1.9383478  1.3150617 -1.4739595  .14049248     _Iyear_83 | -2.4862166   1.433189 -1.7347444  .08278609     _Iyear_85 |  -2.293721  1.6709237 -1.3727264  .16983742     _Iyear_87 | -2.8867738  1.9100228 -1.5113819  .13069118      diff-int | -2.9745408  2.2316307 -1.3329001   .1825646      diff-age |  .01050808  .02053247  .51177844  .60880608    diff-grade |  .01403913  .02483142  .56537754    .571817    diff-south | -.01017179  .12702618 -.08007635  .93617653 diff-not_smsa | -.24502608  .14435482 -1.6973876  .08962341diff-_Iyear_72 |  4.5353507  2.3909244  1.8969026  .05784079diff-_Iyear_73 |   3.213293  2.1675763  1.4824359  .13822434diff-_Iyear_77 |  2.9858792  2.1187489  1.4092653  .15875674diff-_Iyear_78 |  2.8557536  2.1441469  1.3318834  .18289852diff-_Iyear_80 |  3.4787453  2.1165322  1.6436061  .10025759diff-_Iyear_82 |  2.3113123   2.108686  1.0960913  .27303883diff-_Iyear_83 |  2.4132524  2.1023472  1.1478848  .25101613diff-_Iyear_85 |  2.7441521  2.0885294   1.313916  .18887448diff-_Iyear_87 |  2.6838554   2.079152  1.2908414  .19675869         y-lag |  1.5646588  .06439017  24.299654          0

另外两个命令应用为：

. xi: cquadpseudo union idcode age grade south not_smsa i.yeari.year            _Iyear_70-88        (naturally coded; _Iyear_70 omitted)Fit Pseudo Conditional Maximum Likelihood estimator for the dynamic logit modelsee Bartolucci & Nigro (2012), J.EconometricsFirst step      |      lk       lk-lk0------+--------------------------------------------------    1 | -4550.1859  1.000e+10    2 | -4508.4587  41.727174    3 | -4479.6267  28.832058    4 | -4464.3395  15.287228    5 | -4462.0772  2.2622144    6 |  -4462.077   .0002431    7 |  -4462.077  1.182e-11Second step      |      lk       lk-lk0------+--------------------------------------------------    1 | -3386.3831  1.000e+10    2 | -3072.2352  314.14795    3 | -3068.2783  3.9568752    4 | -3068.2768  .00144833    5 | -3068.2768  5.707e-10             |      est.     s.e.(rob)    t-stat.    p-value-------------+--------------------------------------------------         age |  .18590097   .1250259  1.4868997   .1370413       grade | -.03115066  .05488738 -.56753782  .57034884       south | -.62116171  .16083689 -3.8620598  .00011244    not_smsa |  .10764683  .14923884  .72130574  .47072142   _Iyear_71 |  .66895192  1.0824889  .61797577  .53659131   _Iyear_72 |  .26741545  .96467039  .27720914  .78161952   _Iyear_73 |  .04473093  .83125233   .0538115  .95708535   _Iyear_77 | -.66439033  .34745141  -1.912182  .05585286   _Iyear_78 | -.56283602  .22525065 -2.4987098  .01246463   _Iyear_80 | -.42448135  .08815096 -4.8153911  1.469e-06   _Iyear_82 | -1.3962766  .30057821 -4.6453022  3.396e-06   _Iyear_83 | -1.8777382  .42387865 -4.4298957  9.428e-06   _Iyear_85 | -1.7545693  .66528615 -2.6373152  .00835651   _Iyear_87 |  -2.409943  .91783005 -2.6256963  .00864719   _Iyear_88 | -2.5102739  1.0890816 -2.3049456  .02116962       y-lag |  1.6295114  .07720721  21.105691          0. xi: cquadequ union idcode age grade south not_smsa i.yeari.year            _Iyear_70-88        (naturally coded; _Iyear_70 omitted)Fit quadratic exponential model by Conditional Maximum Likelihoodwith modified interaction termssee Bartolucci, Nigro & Pigini (2013), MPRA Paper 48233      |      lk       lk-lk0------+--------------------------------------------------    1 | -3439.9096  1.000e+10    2 | -3078.6649  361.24465    3 | -3076.1717  2.4932592    4 | -3076.1712  .00049932    5 | -3076.1712  5.093e-11             |      est.      s.e.      t-stat.    p-value-------------+--------------------------------------------------         age |  .16845566  .11901965  1.4153601  .15696294       grade | -.03958659  .04550678 -.86990548  .38435206       south | -.53406297  .13625918  -3.919464  .00008875    not_smsa |   .0984639  .13080979  .75272577  .45161472   _Iyear_71 |  1.6032853  1.0337023  1.5510126  .12089866   _Iyear_72 |  1.1740137  .91650676  1.2809657  .20020572   _Iyear_73 |  .97015581  .79589985  1.2189421  .22286618   _Iyear_77 |  .24177005  .33043231  .73167798  .46436514   _Iyear_78 |  .25282926  .21264697  1.1889624  .23445446   _Iyear_80 |  .54363568  .08483378  6.4082453  1.472e-10   _Iyear_82 |  -.3246461   .2861711 -1.1344475  .25660686   _Iyear_83 | -.88650878  .40228033  -2.203709  .02754482   _Iyear_85 | -.68779397  .63653421 -1.0805295  .27990647   _Iyear_87 | -1.3316314  .87497451 -1.5219087  .12803195   _Iyear_88 | -1.5551096  1.0362781 -1.5006681  .13344142       y-lag |  .76891417  .03180295  24.177448          0.