Panel FMOLS procedure

[Ramiro1969]

Hello everyone, hello Tom.

I am estimating a cointegration vector with the FMOLS procedure for panel data (@PANELFM) and I got the following results:

Code: Select all

@panelfm(det=trend, tdum, average=simple, lags=3, print=short)
#vd alfa beta gama delta

Mean Group Panel FM Estimation
       LHS Variable NO
       Individuals              19
       Time Periods             30
       Kernel Width              3
       common time dummies included

       RHS Variable H0 Coefficient
       ALFA               0.000000
       BETA               0.000000
       GAMA               0.000000
       DELTA              0.000000

Member Variable     Coefficient    t-Statistic
Group  ALFA              -0.004321    0.631423
       BETA              -0.000033   -0.203121
       GAMA              -3.123302   -4.423334
       DELTA              1.209476    3.132785

Can any of you tell me if is normal to get opposite signs in the Coefficient and in the T statistic; such as the one that I am obtaining in the
ALFA variable of my results where: ALFA= -0.004321 (negative sign) and its T statistic = 0.631423 (positive sign).

I was reading Pedroni's classic papers and Tom's explanations, and I understand that it is possible because as Tom says:
"group mean FMOLS is an average of estimators which themselves aren't least squares"
Then, this is correct?
Or am I doing something wrong in the FMOLS estimation and for that reason I have opposite signs in the coefficient and in the T statistic?

Please, help me

Best regards
Ramiro

[TomDoan]

Both the coefficients and the t-stats are averages across sectors. If a set of coefficients takes both signs across sectors, the average t-stat could take a different sign than the average coefficient.

[Ramiro1969]

Both the coefficients and the t-stats are averages across sectors. If a set of coefficients takes both signs across sectors, the average t-stat could take a different sign than the average coefficient.

Thank you Tom

Best regards
Ramiro