data("sexratio")
model <- SSModel(Male ~ SSMtrend(1, Q = NA), u = sexratio[, "Total"],
  data = sexratio, distribution = "binomial")
fit <- fitSSM(model, inits = -15, method = "BFGS")
fit$model["Q"]

# Computing confidence intervals in response scale
# Uses importance sampling on response scale (400 samples with antithetics)

pred <- predict(fit$model, type = "response", interval = "conf", nsim = 100)

ts.plot(cbind(model$y/model$u, pred), col = c(1, 2, 3, 3), lty = c(1, 1, 2, 2))

## Not run: 
# Now with sex ratio instead of the probabilities:
imp <- importanceSSM(fit$model, nsim = 1000, antithetics = TRUE)
sexratio.smooth <- numeric(length(model$y))
sexratio.ci <- matrix(0, length(model$y), 2)
w <- imp$w/sum(imp$w)
for(i in 1:length(model$y)){
 sexr <- exp(imp$sample[i, 1, ])
 sexratio.smooth[i] <- sum(sexr*w)
 oo <- order(sexr)
 sexratio.ci[i, ] <- c(sexr[oo][which.min(abs(cumsum(w[oo]) - 0.05))],
                      sexr[oo][which.min(abs(cumsum(w[oo]) - 0.95))])
}

# Same by direct transformation:
out <- KFS(fit$model, smoothing = "signal", nsim = 1000)
sexratio.smooth2 <- exp(out$thetahat)
sexratio.ci2 <- exp(c(out$thetahat) + qnorm(0.025) *
  sqrt(drop(out$V_theta))%o%c(1, -1))

ts.plot(cbind(sexratio.smooth, sexratio.ci, sexratio.smooth2, sexratio.ci2),
        col = c(1, 1, 1, 2, 2, 2), lty = c(1, 2, 2, 1, 2, 2))

## End(Not run)