Hurlin 的PSTR模型包，怎样确定位置参数个数

http://www.runmycode.org/companion/view/2564
我用的是这个包

1.怎样确定位置参数个数？

2. 门限变量也是解释变量时，结果中（结果如下）的warning 是什么意思？

3. 这个包中是否能够把门限变量也作为解释变量？

我跑出的结果是这样的：其中红字部分是什么？
PLEASE CITE:

Fouquau J., Hurlin C. et Rabaud I. (2008), The Feldstein-Horioka Puzzle: a Panel Smooth Transition Regression Approach
Economic Modelling, vol. 25(2), pp. 284-299

AND

Colletaz G. et Hurlin C. (2006), Threshold Effects in the Public Capital Productivity: An International Panel Smooth Transition Approach
Document de Recherche LEO

Thank you


***********************
*** LINEARITY Tests ***
***********************

H0: Linear Model H1: PSTR model with at least one Threshold Variable (r=1)

     Wald Tests (LM):         W = 26.939    pvalue = 0.000

     Fisher Tests (LMF):      F = 7.245    pvalue = 0.000

     LRT Tests (LRT):         LRT = 28.702    pvalue = 0.000



**************************************************************************
*** TESTING THE NUMBER OF REGIMES: TESTS OF NO REMAINING NON-LINEARITY ***
**************************************************************************

  Initial Conditions : Assumed Number of Thresholds r = 1   Number of Regressions = 270
Initial Conditions on (c,gamma)
    5.0000    4.3856

  Estimation of the Model with r = 1 and m = 1 : Convergence = 1   RSS = 0.054
  RSS under H1 = 0.050

  Initial Conditions : Assumed Number of Thresholds r = 2   Number of Regressions = 270
Initial Conditions on (c,gamma)
    6.5060    5.0000    4.3510    5.1626

  Estimation of the Model with r = 2 and m = 1 : Convergence = 1   RSS = 0.037
  RSS under H1 = 0.035

H0: PSTR with r = 1  against  H1: PSTR with at least r = 2

     Wald Tests (LM):         W = 15.987    pvalue = 0.003

     Fisher Tests (LMF):      F = 3.920    pvalue = 0.004

     LRT Tests (LRT):         LRT = 16.587    pvalue = 0.002

H0: PSTR with r = 2  against  H1: PSTR with at least r = 3

     Wald Tests (LM):         W = 11.522    pvalue = 0.021

     Fisher Tests (LMF):      F = 2.711    pvalue = 0.031

     LRT Tests (LRT):         LRT = 11.829    pvalue = 0.019


Given the choices of rmax = 2 and m = 1, the OPTIMAL (LMF criterion) NUMBER OF THRESHOLD FUNCTIONS is r = 2


**************************************
*** FINAL ESTIMATION OF PSTR MODEL ***
**************************************

   Final Estimation of the Model with r = 2 and m = 1 by NLS ***

   Initial Conditions on (gamma,c) :
    3.2389   10.9404    4.8463    5.1317

WARNING: at least one explicative variable is used as threshold variable （这是什么意思？）

  RSS = 0.037      Convergence = 1   

  AIC = -8.478      BIC = -8.234   

Estimated slope parameter of the transition function (one for for each transition function)
    3.2389   10.9404

Estimated location parameters (per column for each transition function)
    4.8463    5.1317

Estimated slope parameters (per column for each transition function)
    0.0027   -0.0248    0.0227
    0.0141   -0.0796    0.0888
    1.0235    0.0288   -0.0309
    0.0455    0.0540   -0.0701

Standard Errors of estimated slope parameters corrected fo heteroskedasticity (per column for each transition function)
    0.0006    0.0046    0.0032
    0.0051    0.0166    0.0141
    0.0596    0.0147    0.0134
    0.0053    0.0392    0.0385

t-statistics based on corrected standard errors (per column for each transition function)
    4.5518   -5.4346    7.1491
    2.7843   -4.7910    6.3022
   17.1793    1.9629   -2.3015
    8.6537    1.3769   -1.8184

OUPUT: Individual Elasticities for each explicative variable (first column is the cross section identifier) are available in the excel file result1.xls

可以私聊

还有一个问题
当r=3时
为什么位置参数的值会出现 c1>c2<c3的情况，不应该是是c1<c2<c3吗？
这样的的结果该怎么解释？

PLEASE CITE:

Fouquau J., Hurlin C. et Rabaud I. (2008), The Feldstein-Horioka Puzzle: a Panel Smooth Transition Regression Approach
Economic Modelling, vol. 25(2), pp. 284-299

AND

Colletaz G. et Hurlin C. (2006), Threshold Effects in the Public Capital Productivity: An International Panel Smooth Transition Approach
Document de Recherche LEO

Thank you


***********************
*** LINEARITY Tests ***
***********************

H0: Linear Model H1: PSTR model with at least one Threshold Variable (r=1)

     Wald Tests (LM):         W = 100.807    pvalue = 0.000

     Fisher Tests (LMF):      F = 25.507    pvalue = 0.000

     LRT Tests (LRT):         LRT = 114.793    pvalue = 0.000



**************************************************************************
*** TESTING THE NUMBER OF REGIMES: TESTS OF NO REMAINING NON-LINEARITY ***
**************************************************************************

  Initial Conditions : Assumed Number of Thresholds r = 1   Number of Regressions = 270
Initial Conditions on (c,gamma)
    4.0000   16.5500

  Estimation of the Model with r = 1 and m = 1 : Convergence = 1   RSS = 0.277
  RSS under H1 = 0.264

  Initial Conditions : Assumed Number of Thresholds r = 2   Number of Regressions = 270
Initial Conditions on (c,gamma)
    4.2623    5.0000   16.3937    9.1500

  Estimation of the Model with r = 2 and m = 1 : Convergence = 1   RSS = 0.233

WARNING: At least one estimated Location Parameter is outside the trimming for a PTR model

The location Parameter should range from  9.1500  to  20.7500 in a PTR model

  RSS under H1 = 0.224

  Initial Conditions : Assumed Number of Thresholds r = 3   Number of Regressions = 270
Initial Conditions on (c,gamma)
    6.6727   14.4274    1.4000   16.5874    8.9534   18.2100

  Estimation of the Model with r = 3 and m = 1 : Convergence = 1   RSS = 0.206

WARNING: At least one estimated Location Parameter is outside the trimming for a PTR model

The location Parameter should range from  9.1500  to  20.7500 in a PTR model

  RSS under H1 = 0.204

H0: PSTR with r = 1  against  H1: PSTR with at least r = 2

     Wald Tests (LM):         W = 20.259    pvalue = 0.001

     Fisher Tests (LMF):      F = 4.025    pvalue = 0.001

     LRT Tests (LRT):         LRT = 20.749    pvalue = 0.001

H0: PSTR with r = 2  against  H1: PSTR with at least r = 3

     Wald Tests (LM):         W = 16.735    pvalue = 0.005

     Fisher Tests (LMF):      F = 3.256    pvalue = 0.007

     LRT Tests (LRT):         LRT = 17.068    pvalue = 0.004

H0: PSTR with r = 3  against  H1: PSTR with at least r = 4

     Wald Tests (LM):         W = 4.246    pvalue = 0.515

     Fisher Tests (LMF):      F = 0.792    pvalue = 0.556

     LRT Tests (LRT):         LRT = 4.267    pvalue = 0.512


Given the choices of rmax = 3 and m = 1, the OPTIMAL (LMF criterion) NUMBER OF THRESHOLD FUNCTIONS is r = 3


**************************************
*** FINAL ESTIMATION OF PSTR MODEL ***
**************************************

   Final Estimation of the Model with r = 3 and m = 1 by NLS ***

   Initial Conditions on (gamma,c) :
    3.1033   13.9345    1.7756   16.7170    8.9645   23.7064


WARNING FINAL ESTIMATION: At least one estimated Location Parameter is outside the trimming for a PTR model

The location Parameter must range from  9.1500  to  20.7500 for a PTR model

WARNING: at least one explicative variable is used as threshold variable

  RSS = 0.206      Convergence = 1   

  AIC = -7.462      BIC = -7.217   

Estimated slope parameter of the transition function (one for for each transition function)
    3.1033   13.9344    1.7756

Estimated location parameters (per column for each transition function)
   16.7170    8.9645   23.7064

Estimated slope parameters (per column for each transition function)
    0.0833   -0.0152   -0.1035    0.0033
   -0.0347    0.0630    0.0534   -0.0139
    0.3304   -0.0218   -0.3129    1.7043
    1.0586   -0.0353    0.0834   -0.6643
   -0.0586    0.0226    0.0586   -0.0134

Standard Errors of estimated slope parameters corrected fo heteroskedasticity (per column for each transition function)
    0.0262    0.0114    0.0267    0.0125
    0.0100    0.0098    0.0102    0.0151
    0.3504    0.0149    0.3491    0.4974
    0.1087    0.0054    0.1146    0.2006
    0.0244    0.0031    0.0243    0.0024

t-statistics based on corrected standard errors (per column for each transition function)
    3.1743   -1.3382   -3.8816    0.2681
   -3.4805    6.4124    5.2344   -0.9261
    0.9430   -1.4595   -0.8964    3.4263
    9.7404   -6.5173    0.7278   -3.3122
   -2.4072    7.3059    2.4110   -5.5755

c值不是人为确定的吗，一般情况下都是1或2，你设置为3怎么解释呢？还有将门槛设置为解释变量，在将这列数重新命名作为解释变量添加进去应该可以吧。用MATLAB做出来是可以的。
有个问题想请教，这个包可以自动画图吗？画几个图？

您好，如果您的求助没有解决，请到项目交易发布需求，会有更快更专业的用户帮助您https://bbs.pinggu.org/z_prj.php

哈密瓜啊啊啊 发表于 2018-11-24 16:45
c值不是人为确定的吗，一般情况下都是1或2，你设置为3怎么解释呢？还有将门槛设置为解释变量，在将这列数重 ...

如果有3个转换函数，如果C值设为1时，三个转换函数各有一个c值，。
我试过将转换变量重新添加进解释变量，结果是一样的。
我这个包不能自动画图。

我用的这个包是可以的。
一般情况下，转换函数的个数和位置参数的个数是一样的。转换函数为1时对应一个高区制和一个低区制，此时突变点只有一个，也就是位置参数。看得文章不多，没有见过研究会设置成不同的值。

哈密瓜啊啊啊 发表于 2018-11-27 15:57
我用的这个包是可以的。
一般情况下，转换函数的个数和位置参数的个数是一样的。转换函数为1时对应一个高区 ...

WARNING: at least one explicative variable is used as threshold variable
即使把门限变量重新输入到excel 列里，仍然会出现以上报错。
门限值检验帮了我大忙，我以前那个code没有门限值个数检验。画图也很方便

楚少伯 发表于 2018-11-27 20:29
WARNING: at least one explicative variable is used as threshold variable
即使把门限变量重新输入 ...

哈哈，谢谢你的论坛币啦

哈密瓜啊啊啊 发表于 2018-12-4 21:34
哈哈，谢谢你的论坛币啦

还有个问题啊，如果出现三个门限值ABC，A>B, B<C 那该怎么解释？门限值不应该越往后越大？对应的求出的参数是否与门限值所在列相对应？。大神啊，如果能解答我的问题，我把你的答案设为最佳