diff --git a/R/sparse.R b/R/sparse.R
index 92b45d3..d588e30 100644
--- a/R/sparse.R
+++ b/R/sparse.R
@@ -2,6 +2,13 @@
 #' @param obj The TMB object with random effects turned on and
 #'   optimized
 #' @param iter Total iterations to run (warmup + sampling)
+#' @param init Either 'last.par.best' (default) or 'random'. The
+#'   former starts from the joint mode and the latter draws from
+#'   a multivariate t distribution with df=1 centered at the mode
+#'   using the inverse joint precision matrix as a covariance
+#'   matrix. Note that StanEstimators only allows for the same
+#'   init vector for all chains currently. If a seed is specified
+#'   it will be set and thus the inits used will be reproducible.
 #' @param warmup Total warmup iterations
 #' @param chains Number of chains
 #' @param cores Number of parallel cores to use
diff --git a/man/pairs_admb.Rd b/man/pairs_admb.Rd
index 4fa1d45..c53c9dd 100644
--- a/man/pairs_admb.Rd
+++ b/man/pairs_admb.Rd
@@ -8,6 +8,7 @@ confidence ellipses.}
 pairs_admb(
   fit,
   order = NULL,
+  inc_warmup = FALSE,
   diag = c("trace", "acf", "hist"),
   acf.ylim = c(-1, 1),
   ymult = NULL,
@@ -30,6 +31,9 @@ NULL (default) to use the order declared in the model, or
 sizes ordered by slowest or fastest mixing respectively. See
 example for usage.}
 
+\item{inc_warmup}{Whether to include the warmup samples or not
+(default).}
+
 \item{diag}{What type of plot to include on the diagonal,
 options are 'acf' which plots the autocorrelation function
 \code{acf}, 'hist' shows marginal posterior histograms, and
diff --git a/man/sample_sparse_tmb.Rd b/man/sample_sparse_tmb.Rd
index 0dd8ad9..a790142 100644
--- a/man/sample_sparse_tmb.Rd
+++ b/man/sample_sparse_tmb.Rd
@@ -11,7 +11,9 @@ sample_sparse_tmb(
   cores,
   chains,
   control = NULL,
-  seed = NULL
+  seed = NULL,
+  init = c("last.par.best", "random"),
+  metric = c("sparse", "dense", "diag", "unit")
 )
 }
 \arguments{
@@ -29,6 +31,14 @@ optimized}
 \item{control}{NUTS control list}
 
 \item{seed}{Random number seed}
+
+\item{init}{Either 'last.par.best' (default) or 'random'. The
+former starts from the joint mode and the latter draws from
+a multivariate t distribution with df=1 centered at the mode
+using the inverse joint precision matrix as a covariance
+matrix. Note that StanEstimators only allows for the same
+init vector for all chains currently. If a seed is specified
+it will be set and thus the inits used will be reproducible.}
 }
 \value{
 A fitted MCMC object of class 'adfit'