Seasonal State Space Model

[cherry]

Hello.
Can you explain me what's wrong in my specification of quarterly state space model?

I used the following code:

param c(1) 0 c(2) 0 c(3) 0 c(4) 0
@signal log(ukgas) = mu+gamma+[var=exp(c(1))]
@state mu= mu(-1)+nu(-1)+[var=exp(c(2))]
@state nu= nu(-1)+[var=exp(c(3))]
@state gamma = -gamma(-1)-gamma2(-1)-gamma3(-1)+[var=exp(c(4))]
@state gamma2=gamma(-1)+[var=0]
@state gamma3=gamma2(-1)+[var=0]

The result is, however, discouraging:

Sspace: SSP1
Method: Maximum likelihood (Marquardt)
Date: 06/21/10 Time: 22:22
Sample: 1960Q1 2000Q4
Included observations: 164
Valid observations: 108
Estimation settings: tol= 0.00010, derivs=accurate numeric
Initial Values: C(1)=0.00000, C(2)=0.00000, C(3)=0.00000, C(4)=0.00000
Convergence achieved after 42 iterations
WARNING: Singular covariance - coefficients are not unique

Coefficient Std. Error z-Statistic Prob.

C(1) -6.307193 NA NA NA
C(2) -346.4645 NA NA NA
C(3) -11.74829 NA NA NA
C(4) -5.711534 NA NA NA

Final State Root MSE z-Statistic Prob.

MU 7.931120 0.807107 9.826601 0.0000
NU 0.024651 0.022359 1.102469 0.2703
GAMMA 0.615757 0.314427 1.958347 0.0502
GAMMA2 0.144682 0.306988 0.471294 0.6374
GAMMA3 -0.680491 0.306457 -2.220508 0.0264

Log likelihood 44.65386 Akaike info criterion -0.752849
Parameters 4 Schwarz criterion -0.653511
Diffuse priors 5 Hannan-Quinn criter. -0.712571

What is the right way to write seasonal coefficients equation?