连老师:
您好!
我最近在做面板门限模型,采用的是xtptm命令,对回归结果不是很清楚,想请教一下您,谢谢您!
以下是存在单门限效应后,进行的双门限效应检验的结果,问题有蓝色字体注明。
================ Fundamental information: ===================
Number of Regime independent variables: 4
Number of Regime dependent variables: 1
Number of individuals in panel: 15
Number of periods in panel: 6
================ Ordinary Fixed effect regression: =========
Sum of Squared Residuls: 0.0747
Stardard error of regression: 0.0327
Regression Result(ordinary standard error):
Coef Std t Prob
1 -0.0382 0.0103 -3.7041 0.0004
2 4.1930 0.6816 6.1517 0.0000
3 -0.0015 0.0018 -0.8334 0.4075
4 -0.0005 0.0015 -0.3480 0.7289
5 -0.0257 0.0068 -3.7504 0.0004
Regression Result(Robust standard error):
Coef Std_Robust t Prob
1 -0.0382 0.0100 -3.8170 0.0003
2 4.1930 0.8608 4.8709 0.0000
3 -0.0015 0.0012 -1.2798 0.2048
4 -0.0005 0.0012 -0.4203 0.6755
5 -0.0257 0.0056 -4.5611 0.0000
================ Single threshold regession: ===================
Minimized Sum of Squared Residuals: 0.0705
Standard error of residuals: 0.0320
Threshold estimator: 6.6300
95% conf. intv. of threshold: 4.1000 17.5000
LR Critical value to test gamma=gamma0: 7.3523 此次只给出了LR的临界值,如何获得LR?
Threshold regression(Ordinary Std. Error):
Coef Std t prob
1 -0.0324 0.0105 -3.0884 0.0029
2 4.1087 0.6682 6.1489 0.0000
3 -0.0008 0.0018 -0.4543 0.6510
4 -0.0006 0.0014 -0.4176 0.6775
5 -0.0181 0.0077 -2.3540 0.0214
6 -0.0302 0.0071 -4.2796 0.0001
Threshold regression(Robust Std. Error):
Coef Std_Robust t prob
1 -0.0324 0.0109 -2.9781 0.0040
2 4.1087 0.8260 4.9744 0.0000
3 -0.0008 0.0013 -0.6301 0.5307
4 -0.0006 0.0012 -0.5213 0.6038
5 -0.0181 0.0062 -2.9213 0.0047
6 -0.0302 0.0060 -5.0328 0.0000
Note: Critcal (Inverse CDF of LR stat): -2*ln(1-sqrt(1-alpha))
Ho: No threshold; Ha: Single threshold
Number of bootstrap:10000
F-stat & Prob: 4.1149 0.0482 此次的F值检验是LM检验吗?说明6.63的门限值是正确的吗?那跟LR检验有何区别?
F-critical value of 90% 95% 99%:
1
+---------------+
1 2.802021258
2 4.041741762
3 7.040245107
+---------------+
Thresholds in single model:
6.630000114
============ Double threshold regession: stage = 1
Minimized Sum of Squared Residuals: 0.0609
Standard error of residuals: 0.0299
Threshold estimate: 8.9000
95% conf. int. of threshold: 8.6000 9.9700
Two Thresholds: 6.6300 8.9000
Threshold regression(Ordinary Std. Err.):
Coef Std t prob
1 -0.0323 0.0098 -3.2903 0.0016
2 4.2186 0.6267 6.7314 0.0000
3 -0.0009 0.0017 -0.5248 0.6014
4 -0.0002 0.0014 -0.1500 0.8812
5 -0.0272 0.0077 -3.5245 0.0008
6 -0.0492 0.0088 -5.5883 0.0000
7 -0.0291 0.0066 -4.3874 0.0000
Threshold regression(Robust Std. Err.):
Coef Std_Robust t prob
1 -0.0323 0.0103 -3.1417 0.0025
2 4.2186 0.7129 5.9176 0.0000
3 -0.0009 0.0010 -0.9210 0.3603
4 -0.0002 0.0011 -0.1832 0.8552
5 -0.0272 0.0066 -4.1013 0.0001
6 -0.0492 0.0079 -6.2570 0.0000
7 -0.0291 0.0057 -5.1304 0.0000
============ Double threshold regession: stage = 2
Minimized Sum of Squared Residuals: 0.0609
Standard error of residuals: 0.0299
Threshold estimate: 6.6300
95% conf. int. of threshold: 6.2000 8.0000
Two Thresholds: 6.6300 8.9000
Threshold regression(Ordinary Std. Err.):
Coef Std t prob
1 -0.0323 0.0098 -3.2903 0.0016
2 4.2186 0.6267 6.7314 0.0000
3 -0.0009 0.0017 -0.5248 0.6014
4 -0.0002 0.0014 -0.1500 0.8812
5 -0.0272 0.0077 -3.5245 0.0008
6 -0.0492 0.0088 -5.5883 0.0000
7 -0.0291 0.0066 -4.3874 0.0000
Threshold regression(Robust Std. Err.):
Coef Std_Robust t prob
1 -0.0323 0.0103 -3.1417 0.0025
2 4.2186 0.7129 5.9176 0.0000
3 -0.0009 0.0010 -0.9210 0.3603
4 -0.0002 0.0011 -0.1832 0.8552
5 -0.0272 0.0066 -4.1013 0.0001
6 -0.0492 0.0079 -6.2570 0.0000
7 -0.0291 0.0057 -5.1304 0.0000
Thresholds in double model:
1
+---------------+
1 8.899999619
2 6.630000114
+---------------+
Ho: Single threshold; Ha: Double threshold
Number of bootstrap:10000
F-stat & Prob: 10.6670 0.0026 说明双门限效应显著,是否还要进行似然比检验来证明门限值的正确?如果需要的话,如何检验?谢谢?
F-critical value of 90% 95% 99%:
1
+---------------+
1 2.817670082
2 4.034328753
3 7.289280711
+---------------+
============== Descrpitive statistic in each regime: ===========
Descriptive statistics of y-X-THR at regime : 1
Mean Std Min Max Count
1 0.6059 0.0563 0.5119 0.6970 25
2 13.3074 0.4465 12.3110 14.2677 25
3 0.0159 0.0056 0.0077 0.0281 25
4 -4.2534 4.2155 -19.5177 -1.2384 25
5 40.8236 4.1407 31.8100 48.5000 25
6 1.7287 0.6258 1.2320 2.7440 25
7 4.7048 1.9024 -1.5000 6.6000 25
Descriptive statistics of y-X-THR at regime : 2
Mean Std Min Max Count
1 0.5322 0.0573 0.4119 0.6201 24
2 13.8660 1.1822 10.7907 15.6805 24
3 0.0292 0.0363 0.0010 0.1647 24
4 -1.7561 0.7028 -3.3936 -1.0480 24
5 37.9521 4.4123 27.9700 46.2600 24
6 1.4193 0.3923 1.2320 2.7440 24
7 7.8679 0.7284 6.6300 8.8000 24
Descriptive statistics of y-X-THR at regime : 3
Mean Std Min Max Count
1 0.5088 0.0659 0.3904 0.6761 41
2 14.2139 1.7111 10.6556 16.4151 41
3 0.0601 0.0589 0.0010 0.1569 41
4 -1.3330 0.4012 -2.2702 -0.4484 41
5 34.3267 4.6901 23.4000 42.1700 41
6 1.8314 0.6284 1.2320 2.7440 41
7 11.8295 3.6932 8.9000 26.9000 41
LR and Thresholds series are stored in Stata matrix: LR#
You can observe the scatter plot by: _matplot LR#, columns(1 2)
我在命令窗口输入_matplot LR, columns(1 2),得到的散点图,但是横坐标标着C2,纵坐标标注C1,每个点标注着ri(i=1,2……)这代表什么意思?
另外,门限模型的解释变量中可以包括被解释变量的滞后项吗?如果是面板数据,是不是加入滞后项是不是会引起内生性问题?
提到了很多问题,希望能到连老师的解答,谢谢您!