Lead operator F in Panel data regression - As Dependent variable.

[Radhika Channanamchery]

Hi Statalist,

Can a lead operator be used along with the dependent variable in panel data regression?

My study uses land use change analysis using historical panel data.

I am looking to see if last year's change in arable and barren areas would result in the current year's area under non-agriculture as conversion from one category to another and its reporting does not happen together. How do I model this for a panel regression? I have used F(Lead operator) in the DV. Is that right?


I would appreciate any help.

Thanks beforehand

Radhika

 xtreg  F.shreofnonagrlarea shareofbarrenarea  shareofarablearea ,fe

Fixed-effects (within) regression               Number of obs      =       540
Group variable: disid                           Number of groups   =        27

R-sq:  within  = 0.8961                         Obs per group: min =        20
       between = 0.1064                                        avg =      20.0
       overall = 0.1100                                        max =        20

                                                F(2,511)           =   2203.73
corr(u_i, Xb)  = -0.8376                        Prob > F           =    0.0000

-----------------------------------------------------------------------------------
F.                |
shreofnonagrlarea |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
------------------+----------------------------------------------------------------
shareofbarrenarea |  -1.312984   .0765696   -17.15   0.000    -1.463414   -1.162554
shareofarablearea |  -.8670699   .0130697   -66.34   0.000    -.8927469    -.841393
            _cons |   .7183172   .0100656    71.36   0.000     .6985421    .7380923
------------------+----------------------------------------------------------------
          sigma_u |  .16110238
          sigma_e |  .00409206
              rho |  .99935524   (fraction of variance due to u_i)
-----------------------------------------------------------------------------------
F test that all u_i=0:     F(26, 511) =  9189.29             Prob > F = 0.0000

.

.

[George Ford]

Sure, but not sure why you wouldn't just use lags of the X's instead? It's more traditional, but you'll get the same results. I had one case where I had a lot more data on the DV than the X's, so f. was more useful to retain more data.

webuse nlswork , clear
xtset idcode year

xtreg f.ln_w age ttl_exp tenure , fe
xtreg ln_w l.age l.ttl_exp l.tenure , fe

[Radhika Channanamchery]

Sure, but not sure why you wouldn't just use lags of the X's instead? It's more traditional, but you'll get the same results. I had one case where I had a lot more data on the DV than the X's, so f. was more useful to retain more data.

webuse nlswork , clear
xtset idcode year

xtreg f.ln_w age ttl_exp tenure , fe
xtreg ln_w l.age l.ttl_exp l.tenure , fe

Thank you for your kind reply. It helps a lot. I have one more question: When we use lagged independent variables in the panel regression model, do we need to look for autocorrelation? I tried to use the xtserial command to check for this, but it will not accept the time-series lagged response variables, and I used it without lags as below. Is it correct? Do I need to check for any other post-regression tests?
Please help with this. Thank you

 xtreg  shreofnonagrlarea L. shareofbarrenarea L. shareofarablearea,fe

Fixed-effects (within) regression               Number of obs      =       540
Group variable: disid                           Number of groups   =        27

R-sq:  within  = 0.8961                         Obs per group: min =        20
       between = 0.1064                                        avg =      20.0
       overall = 0.1100                                        max =        20

                                                F(2,511)           =   2203.73
corr(u_i, Xb)  = -0.8376                        Prob > F           =    0.0000

-----------------------------------------------------------------------------------
shreofnonagrlarea |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
------------------+----------------------------------------------------------------
shareofbarrenarea |
              L1. |  -1.312984   .0765696   -17.15   0.000    -1.463414   -1.162554
                  |
shareofarablearea |
              L1. |  -.8670699   .0130697   -66.34   0.000    -.8927469    -.841393
                  |
            _cons |   .7183172   .0100656    71.36   0.000     .6985421    .7380923
------------------+----------------------------------------------------------------
          sigma_u |  .16110238
          sigma_e |  .00409206
              rho |  .99935524   (fraction of variance due to u_i)
-----------------------------------------------------------------------------------
F test that all u_i=0:     F(26, 511) =  9189.29             Prob > F = 0.0000

. xttest3

Modified Wald test for groupwise heteroskedasticity
in fixed effect regression model

H0: sigma(i)^2 = sigma^2 for all i

chi2 (27)  =    7.8e+05
Prob>chi2 =      0.0000


. xtserial  shreofnonagrlarea shareofbarrenarea shareofarablearea

Wooldridge test for autocorrelation in panel data
H0: no first order autocorrelation
    F(  1,      26) =    100.164
           Prob > F =      0.0000

. xi:xtreg  shreofnonagrlarea L. shareofbarrenarea L. shareofarablearea,re

Random-effects GLS regression                   Number of obs      =       540
Group variable: disid                           Number of groups   =        27

R-sq:  within  = 0.8961                         Obs per group: min =        20
       between = 0.1065                                        avg =      20.0
       overall = 0.1100                                        max =        20

                                                Wald chi2(2)       =   3950.22
corr(u_i, X)   = 0 (assumed)                    Prob > chi2        =    0.0000

-----------------------------------------------------------------------------------
shreofnonagrlarea |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
------------------+----------------------------------------------------------------
shareofbarrenarea |
              L1. |  -1.285381   .0799252   -16.08   0.000    -1.442032   -1.128731
                  |
shareofarablearea |
              L1. |  -.8533832   .0135878   -62.81   0.000    -.8800148   -.8267516
                  |
            _cons |    .708109   .0211089    33.55   0.000     .6667363    .7494818
------------------+----------------------------------------------------------------
          sigma_u |  .09073038
          sigma_e |  .00409206
              rho |     .99797   (fraction of variance due to u_i)
-----------------------------------------------------------------------------------

. xtoverid

Test of overidentifying restrictions: fixed vs random effects
Cross-section time-series model: xtreg re   
Sargan-Hansen statistic  56.843  Chi-sq(2)    P-value = 0.0000

. xtreg  shreofnonagrlarea L. shareofbarrenarea L. shareofarablearea,fe robust

Fixed-effects (within) regression               Number of obs      =       540
Group variable: disid                           Number of groups   =        27

R-sq:  within  = 0.8961                         Obs per group: min =        20
       between = 0.1064                                        avg =      20.0
       overall = 0.1100                                        max =        20

                                                F(2,26)            =   2268.60
corr(u_i, Xb)  = -0.8376                        Prob > F           =    0.0000

                                      (Std. Err. adjusted for 27 clusters in disid)
-----------------------------------------------------------------------------------
                  |               Robust
shreofnonagrlarea |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
------------------+----------------------------------------------------------------
shareofbarrenarea |
              L1. |  -1.312984    .164084    -8.00   0.000    -1.650264   -.9757047
                  |
shareofarablearea |
              L1. |  -.8670699   .0129038   -67.19   0.000    -.8935941   -.8405457
                  |
            _cons |   .7183172   .0119108    60.31   0.000     .6938343    .7428002
------------------+----------------------------------------------------------------
          sigma_u |  .16110238
          sigma_e |  .00409206
              rho |  .99935524   (fraction of variance due to u_i)
-----------------------------------------------------------------------------------

.

[George Ford]

use clustered standard errors.