Prediction based on bayesian model - Cross Validated

i have created bayesian model estimates 6 parameters using rjags r. want predictions based on new data in r. can me example.

model {    b.m[1] <- kb # mean b[1]    b[1] ~ dnorm( b.m[1], tau_b)    for(t in 2 : 10) { # process model       b.m[t]<- max( b[t-1] + rb[t-1](1 -b[t-1]/k) - c[t-1] , 0.0)       b[t] ~ dnorm(b.m[t], tau_b) } # observation mode       (t in 1: 10) {          logc_mean[t] <- log( max(qsegb[t]eff[t], 0.000001))          c[t] ~ dlnorm(logc_mean[t], tau_c) } # prior          b~dbeta(1, 1)          cv <- 1          k.mean <- 500000          k.v <- pow(cvk.mean, 2)          k.tau <- 1/k.v          k ~ dnorm(k.mean, k.tau)          r ~ dlnorm(0.0, 1.4)          qseg ~ dlnorm(-15.4, 1.44)          tau_c ~ dgamma(0.001,0.001)          tau_b ~ dgamma(0.01,0.01)        }     } 

generally, in bayesian model predictions on new data same way non-bayesian models. example complicated provide simplified 1 make things easier illustrate. want estimate linear regression model

$$ y_i = \beta_0 + \beta_1 x_i + \varepsilon_i $$

and based on model want predict $y_\text{new}$ values given $x_\text{new}$ data. in case plug in $x_\text{new}$ estimated model , jags samples $y_\text{new}$ values based on model. code similar this:

 beta0 ~ dnorm(0, 10)  beta1 ~ dnorm(0, 10)  sigma ~ dunif(0, 50)  (i in 1:n) {     y[i] ~ dnorm(beta0 + beta1 * x[i], sigma)  }  (j in 1:nnew) {     ynew[j] ~ dnorm(beta0 + beta1 * xnew[j], sigma)  } 

where y, x , xnew data vectors , ynew variable storing predictions. distribution of values plausible given estimated model. since model probabilistic, prediction probabilistic, i.e. whole distribution of possible ynew values. point-values take average of ynew, can make prediction intervals taking highest density intervals ynew values.