How to run sampler with sigma fixed?

[jacobenglert]

Suppose we are fitting a normal BART model using a dbarts sampler and we know the true residual variance is equal to one. When I specify resid.prior = fixed(1), it seems like the sampler is respecting my wishes. But when I interrogate the state of the sampler after running there is a sigma $\ne 1$. Can someone explain what exactly is happening here?

I actually do have a problem where I will need to be able to do this. The above solution appears to be working despite the sigma inside sampler$state, but I just want to confirm. If this actually is not working, I will need to find some other way around it.

Reproducible example:

# Fixed sigma reprex

# Simulate Data
set.seed(1)
n <- 100
p <- 5
x <- t(replicate(n, runif(p, 0, 1)))
f <- function(x){
  10*sin(pi*x[,1]*x[,2]) + 20*(x[,3] - .5)^2 + 10*x[,4] + 5*x[,5]
}

y <- rnorm(n, f(x), 1)

# Create dbarts sampler object
library(dbarts)
control <- dbartsControl(n.trees = 100, n.samples = 1000, n.burn = 0,
                         n.chains = 1, keepTrees = TRUE, keepTrainingFits = TRUE,
                         updateState = TRUE, verbose = FALSE)

sampler <- dbarts(x, y, control = control, resid.prior = fixed(1), sigma = 1)

sigma <- numeric(1000)
for(k in seq_len(1000)){
  samples <- sampler$run(0, 1)
  sigma[k] <- samples$sigma
}

# Print sigma stored in state
print(sampler$state[[1]]@sigma)

# Plot sigma values
plot(sigma)

Thanks!