Hello everyone

Dear Joao Santos Silva

I want to estimate a gravity equation on agricultural exports from Country i (a single exporter) to its 76 trading partners over 17 years (2001-2017), 1292 observations (stata 14.2).
The export data matrix contains zero values for 50% of the data.
With the data in my processions:
VAR dep : Xij (agricultural export),
VAR indep : agricultural GDP (PIB) , DISTij, agricultural area (SAU), exchange rate (TxCHG), rural population ((POPRRL), Contiguity dummy (FRONT), Comlangue dummy (LANGUE), trade agreement with EU dummy(ACRue), Agreement with Arab countries dummy (Gzale), and Littoral dummy.

I would like to determine the effects of agricultural PIBi, SAUi and POPiRRL on exports Xij.
I used the xtpoisson, ppml and xtpqml commands as follows with i.idpays fixed effect destination, i.année fixed effect temporel

a) xtpoisson xij lnPIBi lnPIBj lnDISTij lnSAUi lnSAUj lnTxCHG lnPOPiRRL lnPOPjRRL FRONT LANGUE ACRue Gzale Littoral, fe robust

b) xi:ppml xij lnPIBi lnPIBj lnDISTij lnSAUi lnSAUj lnTxCHG lnPOPiRRL lnPOPjRRL FRONT LANGUE ACRue Gzale Littoral i.idpays i.année, cluster (idpays)

c) xtpqml xij lnPIBi lnPIBj lnDISTij lnSAUi lnSAUj lnTxCHG lnPOPiRRL lnPOPjRRL FRONT LANGUE ACRue Gzale Littoral _ Iidpays_* _Iannée_*, cluster (idpays)

the results :

a) xtpoisson xij lnPIBi lnPIBj lnDISTij lnSAUi lnSAUj lnTxCHG lnPOPiRRL lnPOPjRRL FRONT LANGUE ACRue Gzale Littoral, fe robust

note: FRONT dropped because it is constant within group
note: LANGUE dropped because it is constant within group
note: Littoral dropped because it is constant within group

Iteration 0: log pseudolikelihood = -170305.2
Iteration 1: log pseudolikelihood = -170294.46 (not concave)
cannot compute an improvement -- discontinuous region encountered
r(430);

b) xi:ppml xij lnPIBi lnPIBj lnDISTij lnSAUi lnSAUj lnTxCHG lnPOPiRRL lnPOPjRRL FRONT LANGUE ACRue Gzale Littoral i.idpays i.année, cluster (idpays)

i.idpays _Iidpays_1-76 (naturally coded; _Iidpays_1 omitted)
i.année _Iannée_2001-2017 (naturally coded; _Iannée_2001 omitted)

note: checking the existence of the estimates
WARNING: lnPOPjRRL has very large values, consider rescaling or recentering

Number of regressors excluded to ensure that the estimates exist: 0
Number of observations excluded: 0

note: lnSAUi omitted because of collinearity
note: lnPOPiRRL omitted because of collinearity
note: _Iidpays_45 omitted because of collinearity
note: _Iidpays_69 omitted because of collinearity
note: _Iidpays_71 omitted because of collinearity
note: _Iidpays_74 omitted because of collinearity
note: _Iannée_2014 omitted because of collinearity

note: starting ppml estimation

Iteration 1: deviance = 392080.8
Iteration 2: deviance = 259000.8
Iteration 3: deviance = 230799.1
Iteration 4: deviance = 224330.5
Iteration 5: deviance = 222953.6
Iteration 6: deviance = 222715.5
Iteration 7: deviance = 222686.7
Iteration 8: deviance = 222684.1
Iteration 9: deviance = 222684
Iteration 10: deviance = 222684
Iteration 11: deviance = 222684

Number of parameters: 98
Number of observations: 1292
Pseudo log-likelihood: -113084.73
R-squared: .9224159
Option strict is: off
(Std. Err. adjusted for 76 clusters in idpays)

Robust
xij Coef. Std. Err. z P>z [95% Conf. Interval]

lnPIBi 1.078178 .3767022 2.86 0.004 .3398549 1.8165
lnPIBj .5543666 .5529202 1.00 0.316 -.5293372 1.63807
lnDISTij -3.964505 1.640279 -2.42 0.016 -7.179393 -.749616
lnSAUj -.0596616 .50623 -0.12 0.906 -1.051854 .9325309
lnTxCHG .5150196 .4361249 1.18 0.238 -.3397696 1.369809
lnPOPjRRL .0688833 .1651138 0.42 0.677 -.2547337 .3925003
FRONT -.0977321 1.528801 -0.06 0.949 -3.094127 2.898663
LANGUE -1.999878 1.438236 -1.39 0.164 -4.818768 .8190118
ACRue -1.005604 .3751885 -2.68 0.007 -1.74096 -.2702478
Gzale .4757488 .3538078 1.34 0.179 -.2177016 1.169199
Littoral 3.420358 1.793566 1.91 0.057 -.094968 6.935683
_Iidpays_2 -6.805872 3.32712 -2.05 0.041 -13.32691 -.2848359
_Iidpays_3 1.965097 1.906009 1.03 0.303 -1.770611 5.700806
_Iidpays_4 -1.647627 1.640553 -1.00 0.315 -4.863051 1.567797
_Iidpays_5 5.330335 1.23936 4.30 0.000 2.901235 7.75943 .
..
.
.
.
.
.
c) xtpqml xij lnPIBi lnPIBj lnDISTij lnSAUi lnSAUj lnTxCHG lnPOPiRRL lnPOPjRRL FRONT LANGUE ACRue Gzale Littoral _ Iidpays_* _Iannée_*, cluster (idpays)

Iteration 0: log likelihood = -170305.2
Iteration 1: log likelihood = -170298.66 (backed up)
Iteration 2: log likelihood = -170246.49 (backed up)
Iteration 3: log likelihood = -170246.28 (backed up)
discontinuous region encountered
cannot compute an improvement
r(430);



1) Are the commands correct?
2) Why with xtpoisson and xtpqml I can't get the results?
3) some coefficient estimates are not statistically significant (greater than 5%), do I remove them from the equation?

Please I would need your help to be able to complete my research work.

Thank you in advance.
Cordially.