What is the syntax of Panel ARDL model with DCCE estimator (xtdcce2)?

[Wondesen Teshome]

<sub>Dear all,

I am learning Stata and trying to find out how to use the xtdcce2?

I installed xtdcce2 version 2.</sub>

I employed 21 countries and 48 years of annual data.

The Dynamic Common Correlated Effects Estimator employed to estimate using Jackknife bias corrections method.

However, the command keeps showing the following error.

. xtdcce2 lrgdpna L.lrgdpna lhc lemp, cr(lrgdpna lhc lemp) cr_lags(4) res(residuals) jack
m_xtdcce_solver(): 3499 selectindex() not found
xtdcce_m_partialout(): - function returned error
<istmt>: - function returned error

. xtdcce2 lrgdpna L.lrgdpna lhc lemp, cr(lrgdpna lhc lemp) cr_lags(4) res(residuals) rec
m_xtdcce_solver(): 3499 selectindex() not found
xtdcce_m_partialout(): - function returned error
<istmt>: - function returned error

Could you suggest me the way out?

[Justin Niakamal]

xtdcce2 is a user-written command by JanDitzen from Stata Journal. I've not used this command, but it looks like your issue might be with res(residuals).

Try taking that out. residual appears to be a postestimation option that you can use with predict.

xtdcce2 lrgdpna L.lrgdpna lhc lemp, cr(lrgdpna lhc lemp) cr_lags(4)  jack
predict res, residuals

[Wondesen Teshome]

Dear @Justin Blasongame

Still, It doesn't work. I think It is not about post-estimation. The former estimation command is not working by itself.

[Justin Niakamal]

I can't replicate your issue. The dataset is different, but the set up is the same. It might help if you post an example of your dataset using dataex (see the FAQ for details).

net get xtdcce2
sysuse xtdcce2_sample_dataset.dta, clear
xtdcce2 log_rgdpo L.log_rgdpo log_hc log_ck, cr(log_rgdpo  log_hc log_ck) cr_lags(4) jack
xtdcce2 log_rgdpo L.log_rgdpo log_hc log_ck, cr(log_rgdpo  log_hc log_ck) cr_lags(4) rec

. xtdcce2 log_rgdpo L.log_rgdpo log_hc log_ck, cr(log_rgdpo  log_hc log_ck) cr_lags(4) jack
(Dynamic) Common Correlated Effects Estimator - Mean Group

Panel Variable (i): id                           Number of obs     =       3999
Time Variable (t): year                          Number of groups  =         93
                                                 Obs per group (T) =         43
Degrees of freedom per group:
 without cross-sectional averages   = 40         F(1675, 2324)     =       2.41
 with cross-sectional averages      = 24         Prob > F          =       0.00
Number of                                        R-squared         =       0.63
 cross sectional lags               = 4          Adj. R-squared    =       0.37
 variables in mean group regression = 279        Root MSE          =       0.05
 variables partialled out           = 1396
                                                 CD Statistic      =       0.64
                                                    p-value        =     0.5247
-------------------------------------------------------------------------------
      log_rgdpo|     Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
---------------+---------------------------------------------------------------
 Mean Group:   |
    L.log_rgdpo|  .4517701   .0325303   13.89    0.000      .3880118   .5155284
         log_hc| -1.026018   .3368268   -3.05    0.002     -1.686186  -.3658495
         log_ck|  .2310946   .0535985    4.31    0.000      .1260434   .3361458
-------------------------------------------------------------------------------
Mean Group Variables: L.log_rgdpo log_hc log_ck
Cross Sectional Averaged Variables: log_rgdpo log_hc log_ck
Heterogenous constant partialled out. Jackknife bias correction used. 

. xtdcce2 log_rgdpo L.log_rgdpo log_hc log_ck, cr(log_rgdpo  log_hc log_ck) cr_lags(4) rec
(Dynamic) Common Correlated Effects Estimator - Mean Group

Panel Variable (i): id                           Number of obs     =       3906
Time Variable (t): year                          Number of groups  =         93
                                                 Obs per group (T) =         42
Degrees of freedom per group:
 without cross-sectional averages   = 39         F(1675, 2231)     =       2.12
 with cross-sectional averages      = 23         Prob > F          =       0.00
Number of                                        R-squared         =       0.61
 cross sectional lags               = 4          Adj. R-squared    =       0.32
 variables in mean group regression = 279        Root MSE          =       0.06
 variables partialled out           = 1396
                                                 CD Statistic      =       8.46
                                                    p-value        =     0.0000
-------------------------------------------------------------------------------
      log_rgdpo|     Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
---------------+---------------------------------------------------------------
 Mean Group:   |
    L.log_rgdpo|  .2326908   .0255933    9.09    0.000      .1825289   .2828527
         log_hc| -.6037607   .4661117   -1.30    0.195     -1.517323   .3098013
         log_ck|  .2694361   .0541726    4.97    0.000      .1632598   .3756124
-------------------------------------------------------------------------------
Mean Group Variables: L.log_rgdpo log_hc log_ck
Cross Sectional Averaged Variables: log_rgdpo log_hc log_ck
Heterogenous constant partialled out. Recursive mean adjustment used to correct for small sample time series bias.

[JanDitzen]

Wondesen Teshome, which Stata version and version of xtdcce2 do you use? It looks like you use version 14 or earlier, which misses the selectindex() command. xtdcce2 should be able to identify this and has its own version of it.

I will check if this is a general bug and might upload an update to my GitHub if necessary tomorrow.

[Wondesen Teshome]

Dear @JanDitzen,

Thank you for the quick response. I am using Stata version 13 and installed xtdcce2 version 2.
Would you do everything you can, including uploading an update?

[Wondesen Teshome]

Dear @JanDitzen,
Thank you!! Problem solved!!
The problem was related with earlier Stata version I used, which had no selectionindex() commands.

[JanDitzen]

Wondesen Teshome, indeed. Did you change Stata versions?

I just uploaded an updated version of 2.21to my GitHub (see https://github.com/JanDitzen/xtdcce2...ster/xtdcce221), you can install it from

net from https://janditzen.github.io/xtdcce2/

. Hope this helps.

Jan

[Wondesen Teshome]

Dear @JanDitzen,

Based on the commendations you gave me, I installed Stata 15.0 and the problem is already solved. I will also use the additional recommendation while I'm not clear on how could I incorporate (add-in) the GitHub version of 2.21 into my Stata.

THANK YOU!!!

[JanDitzen]

Glad to hear it helped.
You can type the code line from above and then click on xtdcce221.
Best,
Jan

[Wondesen Teshome]

Dear members,


In the case of DCCE estimator (xtdcce2), I got two unclear things:

1. one of the independent variables became stationary after its second difference, I(2). Does this variable characteristic affect the estimator?

2. which one of the commands is appropriate to handle the non-stationary characterstic of the variables?

xtdcce2 lrgdpna L.lrgdpna lhc lrnna lemp, noconstant cr(lrgdpna lhc lrnna lemp) cr_lags(3)

xtdcce2 dlrgdpna L.dlrgdpna ddlhc dlrnna dlemp, reportc cr(dlrgdpna ddlhc dlrnna dlemp) cr_lags(3)

could you suggest to me?