diff --git a/matlab/utilities/dataset/compute_acov.m b/matlab/utilities/dataset/nanautocovariance.m
similarity index 57%
rename from matlab/utilities/dataset/compute_acov.m
rename to matlab/utilities/dataset/nanautocovariance.m
index b1834e854..715ae957f 100644
--- a/matlab/utilities/dataset/compute_acov.m
+++ b/matlab/utilities/dataset/nanautocovariance.m
@@ -1,4 +1,4 @@
-function dataset_ = compute_acov(dataset_) 
+function autocov = nanautocovariance(data,order)
 % Computes the (multivariate) auto-covariance function of the sample (possibly with missing observations).
 
 %@info:
@@ -8,36 +8,36 @@ function dataset_ = compute_acov(dataset_)
 %!
 %! @strong{Inputs}
 %! @table @var
-%! @item dataset_
-%! Dynare structure describing the dataset, built by @ref{initialize_dataset}
-%! @item nlag
-%! Integer scalar. The maximum number of lags of the autocovariance function.    
+%! @item data
+%! T*N array of real numbers.
+%! @item order
+%! Integer scalar. The maximum number of lags of the autocovariance function.
 %! @end table
 %!
 %! @strong{Outputs}
 %! @table @var
-%! @item dataset_
-%! Dynare structure describing the dataset, built by @ref{initialize_dataset}
+%! @item autocov
+%! A N*N*order array of real numbers.
 %! @end table
-%! 
-%! @strong{This function is called by:} 
+%!
+%! @strong{This function is called by:}
 %! @ref{descriptive_statistics}.
-%! 
+%!
 %! @strong{This function calls:}
-%! @ref{ndim}, @ref{compute_cova}, @ref{demean}, @ref{nandemean}.
-%!    
-%! @strong{Remark 1.} On exit, a new field is appended to the structure: @code{dataset_.descriptive.acov} is a 
-%! @tex{n\times n\times p} array (where @tex{n} is the number of observed variables as defined by @code{dataset_.info.nvobs}, 
+%! @ref{ndim}, @ref{nancovariance}, @ref{demean}, @ref{nandemean}.
+%!
+%! @strong{Remark 1.} On exit, a new field is appended to the structure: @code{dataset_.descriptive.acov} is a
+%! @tex{n\times n\times p} array (where @tex{n} is the number of observed variables as defined by @code{dataset_.info.nvobs},
 %! and @tex{n} is the maximum number of lags given by the second input @code{nlag}).
-%!    
-%! @strong{Remark 2.} If @code{dataset_.descriptive.cova} does not exist, the covariance matrix is computed prior to the 
+%!
+%! @strong{Remark 2.} If @code{dataset_.descriptive.cova} does not exist, the covariance matrix is computed prior to the
 %! computation of the auto-covariance function.
 %!
 %! @end deftypefn
 %@eod:
 
-% Copyright (C) 2011-2012 Dynare Team
-%    
+% Copyright (C) 2011-2014 Dynare Team
+%
 % This file is part of Dynare.
 %
 % Dynare is free software: you can redistribute it and/or modify
@@ -53,22 +53,23 @@ function dataset_ = compute_acov(dataset_)
 % You should have received a copy of the GNU General Public License
 % along with Dynare.  If not, see <http://www.gnu.org/licenses/>.
 
-% Original author: stephane DOT adjemian AT univ DASH lemans DOT fr
+n = size(data,2);
+missing = isanynan(data);
 
-if ~isfield(dataset_.descriptive,'cova')
-    dataset_ = compute_cova(dataset_);
-end
-dataset_.descriptive.acov = zeros(dataset_.nvobs,dataset_.nvobs,nlag);
+autocov = zeros(n, n, order);
 
-data = transpose(dataset_.data);
-
-for lag=1:nlag
-    for i=1:dataset_.info.nvobs
-        for j=1:dataset_.info.nvobs
-            if dataset_.missing.state
-                dataset_.descriptive.acov(i,j,lag) = nanmean(nandemean(data(lag+1:end,i)).*nandemean(data(1:end-lag,j)));
+for lag=1:order
+    if missing
+        data = nandemean(data);
+    else
+        data = demean(data);
+    end
+    for i=1:n
+        for j=1:n
+            if missing
+                autocov(i,j,lag) = nanmean(data((lag+1):end,i).*data(1:end-lag,j));
             else
-                dataset_.descriptive.acov(i,j,lag) = mean(demean(data(lag+1:end,i)).*demean(data(1:end-lag,j)));
+                autocov(i,j,lag) = mean(data((lag+1):end,i).*data(1:end-lag,j));
             end
         end
     end