Go to the MLX, M, PDF, or HTML version of this file. Go back to fan’s MEconTools Package, Matlab Code Examples Repository (bookdown site), or Math for Econ with Matlab Repository (bookdown site).
We had associative, commutative and distributive laws for scalar algebra, we can think of them as the six bullet points below. Only the multiplicative-commutative law no longer works for matrix, the other rules work for matrix as well as scalar algebra.
Associative laws work as in scalar algebra for matrix
\(\displaystyle (A+B)+C=A+(B+C)\)
\(\displaystyle (A\cdot B)\cdot C=A\cdot (B\cdot C)\)
Commutative Law works as well for addition
\(\displaystyle A+B=B+A\)
with scalars, we know \(3\cdot 4=4\cdot 3\), but commutative law for matrix multiplication does not work, Matrix \(A\cdot B\not= B\cdot A\). The matrix dimensions might not even match up for multiplication. (see below for examples)
And Distributive Law still applies to matrix
\(\displaystyle A\cdot (B+C)=A\cdot B+A\cdot C\)
\(\displaystyle (B+C)\cdot A=B\cdot A+C\cdot A\)
% Non-Square
A = rand(2,3)
A = 2x3
0.6959 0.6385 0.0688
0.6999 0.0336 0.3196
B = rand(3,4)
B = 3x4
0.5309 0.8200 0.5313 0.6110
0.6544 0.7184 0.3251 0.7788
0.4076 0.9686 0.1056 0.4235
% This is OK
disp(A*B)
0.8154 1.0960 0.5847 0.9516
0.5238 0.9076 0.4166 0.5891
% This does not work
try
B*A
catch ME
disp('does not work! Dimension mismatch')
end
does not work! Dimension mismatch
% Square
A = rand(3,3)
A = 3x3
0.0908 0.2810 0.4574
0.2665 0.4401 0.8754
0.1537 0.5271 0.5181
B = rand(3,3)
B = 3x3
0.9436 0.2407 0.6718
0.6377 0.6761 0.6951
0.9577 0.2891 0.0680
% This is OK
A*B
ans = 3x3
0.7030 0.3441 0.2875
1.3704 0.6147 0.5445
0.9773 0.5431 0.5049
% This works, but result differs from A*B
B*A
ans = 3x3
0.2531 0.7252 0.9904
0.3449 0.8432 1.2437
0.1745 0.4322 0.7263In scalar algebra, transpose does not make sense. Given matrix \(A\), \(A^T\) is the transpose matrix of \(A\) where each row of \(A\) becomes columns in \(A^T\). If \(A\) is \(M\) by \(N\), then \(A^T\) is \(N\) by \(M\).
Given matrix \(A\) and scalar value \(r\):
1: \((r\cdot A)^T =r\cdot A^T\)
2: \((A^T )^T =A\)
3: \((A+B)^T =A^T +B^T\)
4: \((A\cdot B)^T =B^T \cdot A^T\)
For the 4th rule, suppose matrix \(A\) is has \(L\) rows and \(M\) columns, and the matrix \(B\) has \(M\) rows and \(N\)columns. \((A\cdot B)\) is a \(L\) by \(N\) matrix, \((A\cdot B)^T\) is a \(N\) by \(L\) matrix. This is equal to \(B^T \cdot A^T\), where we have a \(N\) by \(M\) matrix \(B^T\) multiplied by a \(M\) by \(L\) matrix \(A^T\), and the resulting matrix is \(N\) by \(L\).
A = rand(2,3)
A = 2x3
0.2548 0.6678 0.3445
0.2240 0.8444 0.7805
Atranspose = (A')
Atranspose = 3x2
0.2548 0.2240
0.6678 0.8444
0.3445 0.7805