Remove dqrmvt

Internally rmvt uses both rmvnorm and rchisq. The former is fine, since we were passing in our dqrnorm. However, the latter still uses R's RNG.

In order to use this, we need to implement dqrchisq and mvtnorm::rmvt needs an option rchisq=rchisq.

NEWS.md

@@ -4,7 +4,7 @@
 * Implement weighted sampling with and without replacement.
   Move sampling methods to separate header file, allowing for parallel usage.
   ([#47](https://github.com/daqana/dqrng/pull/47) fixing [#18](https://github.com/daqana/dqrng/issues/18) and [#26](https://github.com/daqana/dqrng/issues/26))
-* New methods `dqrmvnorm` and `dqrmvt` sampling from multivariate normal and t distribution.
+* New methods `dqrmvnorm` sampling from multivariate normal distribution.
   This uses the methods implemented in the `mvtnorm` package and uses `dqrnorm`.
 
 # dqrng 0.3.0

R/dqrmv.R

@@ -24,20 +24,3 @@ dqrmvnorm <- function(n, ...) {
   }
   mvtnorm::rmvnorm(n, ..., rnorm = dqrnorm)
 }
-
-#' @export
-#'
-#' @rdname dqrmv
-#' @examples
-#' ## X ~ t_3(0, diag(2))
-#' x <- dqrmvt(100, sigma = diag(2), df = 3) # t_3(0, diag(2)) sample
-#' plot(x)
-dqrmvt <- function(n, ...) {
-  if (!requireNamespace("mvtnorm", quietly = TRUE)) {
-    stop(
-      "Package \"mvtnorm\" must be installed to use this function.",
-      call. = FALSE
-    )
-  }
-  mvtnorm::rmvt(n, ..., rnorm = dqrnorm)
-}