Subtract combinations of variables in a vector

I have a vector of 7 variables and I need to subtract the all possible combinations of one variable from another. I have tried to loop through the variables but I only get the first variable minus all the rest. I need the loop to continue to subtract variable 2 from the rest. Any help would be greatly appreciated.

for c=1:6; DER1(:,c)=DER(c); DER2(:,c)=DER(c+1);

    for d=1:20;
        DED(:,d) = DER1(c)-DER2(c);
    end
end

Answer 1

[Matt Tearle]

If you have Statistics Toolbox, here's a neat trick (I think this is what you're asking for):

% make a vector of numbers
vec = rand(7,1)
% get inter-point "distances"
dfun = @(x,y) x-y;
pdist(vec,dfun)
% or if you prefer
squareform(pdist(vec,dfun))

EDIT TO ADD

As Teja suggested:

% as a matrix
bsxfun(@minus,vec',vec)
% as a vector
squareform(bsxfun(@minus,vec',vec))

[Teja Muppirala]

This also works:

bsxfun(@minus,vec,vec')

[Teja Muppirala]

But I too like the PDIST function. It is a very handy (and REALLY REALLY FAST) function for getting various distances on multidimensional data.

[Sean de Wolski]

+1 for bsxfun!

[Matt Tearle]

I thought the OP wanted a vector as a result, so I thought pdist was the better candidate. But after thinking about the bsxfun solution, I realized you can always use squareform to go back to a vector. So... (see edit).

Answer 2

[Matt Fig]

Another approach:

S = -diff(nchoosek(vec,2),[],2)

Answer 3

[Paulo Silva]

Here's a version with two for loops

v=1:7;
for a=1:7
    for b=1:7
   %comment the next line if you want the variable to be divided by itself
        if a~=b 
            v(b)=v(b)-v(a);
   %comment the next line if you want the variable to be divided by itself
        end 
    end
end

v

Answer 4

[Andrei Bobrov]

more variant (without Statistics Toolbox)

[I J] = meshgrid(vec);
Mdist = reshape(diff([I(:) J(:)],[],2),[],length(vec));
Vdistt = -Mdist(tril(Mdist)~=0);

MORE variant: first vector, then the matrix

[I J] = meshgrid(vec);
V1 = I(:) - J(:);
[~, b] = unique(abs(V1));
Vdist2 = sortrows([V1(b) b],2);
Mdist2(tril(true(length(vec)),-1)) =  -Vdist2(1:end-1,1);
Mdist2 = Mdist2 - Mdist2.';

[Matt Fig]

I believe the last line should be:

vdistt = -Mdist(tril(true(size(Mdist)),-1));