I am trying to model a state space process using the MARSS package. My
model has two unobservable states and 5 observable time series
along with external covariates in the observation process only. None of the
coefficients in either of the two processes are time varying.
After running my setup and then fitting with MARSS with the fit=FALSE
option turned on I get the following outputs.
Model Structure is
m: 2 state process(es)
n: 5 observation time series
Z : unconstrained
A : time-varying
R : unconstrained
B : unconstrained
U : unconstrained
Q : unconstrained
x0 : fixed
V0 : identity
Obviously I have no idea why A has been classfied as time varying. Can
anyone help me as to how to do this correctly ?
A is initialized like this
A1 <- "zero"
The full code has been copied below
#For two states
B1 <-
matrix(c("b1","b2","b3","b4"),2,2)
#For 2 states
U1 <- matrix(c("u1","u2"),2,1) #For 2 states
Q1 <- "unconstrained"
Z1 <-
matrix(c("z1","z2","z3","z4","z5","z6","z7","z8","z9","z10"),5,2)
#For 2 states
#Initial Parameters
pi1 <- matrix(c(1,2),2,1)
V1 <- diag(1,2)
A1 <- "zero"
R1 <-
matrix(list("r11","r12","r13","r14","r15",
"r12","r22","r23","r24","r25",
"r13","r23","r33","r34","r35",
"r14","r24","r34","r44","r45",
"r15","r25","r35","r45","r55",
"r16","r26","r36","r46","r56")
,5,5)
D <- "unconstrained"
d <- covariate.data
rownames(d) <- rownames(covariate.data)
model.list <-
list(B=B1,U=U1,Q=Q1,Z=Z1,A=A1,R=R1,D=D,d=d,x0=pi1,V0=V1,tinitx=0)
fit <-
MARSS(yt,model=model.list,control=list(maxit=10000,trace=0),fit=FALSE)
Regards
The woods are lovely, dark and deep
But I have promises to keep
And miles to go before I sleep
And miles to go before I sleep
-----
[[alternative HTML version deleted]]