DummyVar.m

function D = dummyvar(group,varargin)
%DUMMYVAR Dummy variable coding.
%   X=DUMMYVAR(GROUP) returns a matrix X containing zeros and ones, whose
%   columns are dummy variables for the grouping variable GROUP.  GROUP can be
%   a categorical or numeric column vector.  The I'th dummy variable column in
%   X contains ones in elements where values in GROUP specify the I'th group.
%
%   The order of the dummy variable columns in X matches the order of the
%   groups defined by GROUP.  When GROUP is a categorical vector, the groups
%   and their order match the output of the CATEGORIES(GROUP) method.   When
%   GROUP is a numeric vector, DUMMYVAR assumes that the groups and their
%   order are 1:MAX(GROUP).  NOTE: In this respect, DUMMYVARS treats a numeric
%   grouping variable differently than GRP2IDX.
%
%   GROUP can also be a cell array containing one or more grouping variables.
%   See "help groupingvariable" for more information.  GROUP can also be a
%   numeric matrix, and DUMMYVARS treats each column as a separate numeric
%   grouping variable, as described above.  With multiple grouping variables,
%   the sets of dummy variable columns are in the same order as the grouping
%   variables in GROUP.
%
%   DUMMYVAR returns a full set of dummy variables for each grouping variable.
%   To create a regression design matrix, you may need to use a smaller set of
%   dummy variables so that SUM(X,2) is not identical to the column in the
%   design matrix that corresponds to the regression intercept.  You can, for
%   example, delete the first or the last dummy variable for each grouping
%   variable.
%
%   Example: Suppose we are studying the effects of two machines and three
%   operators on a process.  The first column of GROUP would have the
%   values one or two depending on which machine was used.  The second
%   column of GROUP would have the values one, two, or three depending on
%   which operator ran the machine.
%
%       machine = [1 1 1 1 2 2 2 2]';
%       oper    = [1 2 3 1 2 3 1 2]';
%       x = dummyvar([machine oper])
%
%   See also GROUPINGVARIABLE.

%   Copyright 1993-2013 The MathWorks, Inc. 




if nargin > 1
    ColumnSkip = varargin{1};
else
    ColumnSkip = 0;
end

if isa(group,'categorical')
    % group by a categorical (nominal or ordinal) variable
    [m,n] = size(group);
    if n~=1
        error(message('stats:dummyvar:BadCateGroup'));
    end
    maxg = length(categories(group));
    group = double(group);
elseif iscell(group)
    % collection of grouping variables in a cell array
    n = numel(group);
    for j=1:n
        gj = group{j};
        gj = grp2idx(gj);
        if j==1
            m = size(gj,1);
            G = zeros(m,n);
        else
            if size(gj,1)~=m
                error(message('stats:dummyvar:InputSizeMismatch'));
            end
        end
        G(:,j) = double(gj);
    end
    group = G;
    maxg = max(group);
else
    % vector or matrix whose columns are grouping variables
    if ~ismatrix(group)
        error(message('stats:dummyvar:BadSizedGroup'));
    end
        
    if min(unique(group)) == 0
        group = group + 1;
    end
    
    [m,n] = size(group);
    if m == 1
        m = n;
        n = 1;
        group = group(:);
    end
    maxg = max(group);
end

if any(any(group - round(group))) ~= 0 || any(any(group<=0))
   error(message('stats:dummyvar:BadGroup'));
end
if ~isfloat(group)
   group = double(group);
end

idxnan = isnan(maxg);
if any(idxnan)
    maxg(idxnan)=0; % set NaN to zero    
    nanstr = sprintf(', %d',find(idxnan));
    warning(message('stats:dummyvar:AllNaNs',nanstr(3:end)));
end

if (ColumnSkip ~= 0) && ((ColumnSkip < 1) || (ColumnSkip > maxg))
    cprintf(rgb('DarkOrange'),'WARNING - index of categories to skeep not consistent with the size of the data (skipping the first category) \n');
    ColumnSkip = 1;
end

colend = cumsum(maxg);          % end col for each var
colidx = [0 colend(1:end-1)];   % for indexing each entry
colstart = 1 + colidx;          % start col
colidx = reshape(colidx(ones(m,1),:),m*n,1);

colD = sum(maxg);
D = zeros(m,colD,class(group));

% Compute linear indices of ones based on row and column numbers
row = (1:m)';
row = reshape(row(:,ones(n,1)),m*n,1);
idx = m*(colidx + group(:) - 1) + row;
D(idx(~isnan(idx))) = 1;

% Use NaN for the columns corresponding to a NaN in the input
for j=1:size(group,2)
    D(isnan(group(:,j)),colstart(j):colend(j)) = NaN;
end

if ColumnSkip ~= 0
    D(:,ColumnSkip) = [];
end