[Gretl-users] Estimation of a GARCH process with conditional skewness and kurtosis

Hi, I am trying to implement a GARCH method which explicitely models the kurtosis and skewness. A short summary could be found here: http://hannes.fedorapeople.org/leon2005.pdf What I got so far is the following: open /usr/share/gretl/data/misc/djclose.gdt series y=100*ldiff(djclose) scalar mu=0.0 scalar beta0=1 scalar beta1=0.4 scalar beta2=0.0 scalar beta3=0.0 scalar gamma0=0.0 scalar gamma1=0.0 scalar gamma2=0.0 scalar delta0=0.0 scalar delta1=0.0 scalar delta2=0.0 mle ll= -0.5*(log(h) - 0.5*(eta)^2 + log(1+(sk/6)*(eta^3-3*eta)+((ku-3)/24)*(eta^4-6*eta^2+3))^2-log(1+((sk^2)/6)+(ku-3)^2/24)) series e=y-mu series h=var(y) series eta=e/sqrt(abs(h)) series sk = (((y - mu)^3) / nobs(y))/ sd(y)^3 series ku = ((((y - mu)^4) / nobs(y))/ sd(y)^4) - 3 series h= beta0+beta1*(e(-1)+ beta3*sqrt(h(-1)))^2+beta2*h(-1) series sk=gamma0 + gamma1*(eta(-1))^3 + gamma2*sk(-1) series ku=delta0 + delta1*(eta(-1))^4 + delta2*ku(-1) params mu beta0 beta1 beta2 beta3 gamma0 gamma1 gamma2 delta0 delta1 delta2 end mle Not sure if everything is right because the execution fails with the message 'Missing values encountered'. I don't really know what's wrong or if my approach is correct. Thanks for any help!