Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I've never really worked with matrices so I would be glad about some information on the following issue:

In some statistic calculations there is often a transposed matrix within a formula. Can someone explain why there is a need to transpose matrices and what properties do they have afterwards?

Here is an example of the Shapiro-Wilk test for normality:

$a_{i} = \left[ (m^T V^{-1} V^{-1} m)^{\frac{1}{2}} \right] m^T V^{-1}$

where $V = \left( \begin{array}{*{3}, c} cov(m_1, m_1) & \cdots & cov(m_1, m_n) \\ \vdots & \ddots & \vdots \\ cov(m_n, m_1) & \cdots & cov(m_n, m_n) \end{array} \right)$

and $m_i = \Phi^{-1} \left( \frac{ i - \frac{3}{8}}{ n + \frac{1}{4}} \right)$

Why is $m_i$ a matrix and why should it be transposed for calculation?

share|cite|improve this question
up vote 3 down vote accepted

Row and column vectors are very often considered as special cases of matrices that happen to have dimension $1$ in one direction. That allows one to write multiplication of matrices and vectors in general form without having to define everything for two cases separately. In your case, $m_i$ is not a matrix, but the column vector $m$ whose coordinates are $m_i$ can be considered as an $n\times1$ matrix.

The reasons for transposing things can be different in different contexts. Often they are related to the fact that the dot product between two vectors $x$ and $y$ can be written as the matrix multiplication $x^Ty$ (it may be instructive to actually draw this if you're not familiar with these things). In your case, $m^TV^{-1}V^{-1}m=m^TV^{-1}{}^TV^{-1}m=(V^{-1}m)^T(V^{-1}m)$ (since $V$ is symmetric and thus equal to its transpose), so that expression is just the dot product of $V^{-1}m$ with itself, i.e. the square of that vector.

share|cite|improve this answer
Thank you, now I can work with the formula! – beyeran Apr 22 '11 at 11:39
@beyeran: as a terminological note, the quantity being rooted in the Shapiro-Wilk test is a quadratic form, which pops up a lot in statistical applications. – J. M. Apr 24 '11 at 15:57
Teeny nitpick: ${V^{-1}}^T$ might benefit from parenthesization, or there's the (you could say awkward) device $V^{-T}$... – J. M. Apr 24 '11 at 16:00

Why do you think that $m_i$ is a matrix? It seems that $m$ is a vector with coordinates $m_i$.

The vector $m$ is transposed, otherwise it could never be multiplied from the left with the matrices, and it is not $a_i=\dots$, it should be $a= \dots$ or $(a_i)=\dots$ as the expression on the right is already a vector and not its components.

share|cite|improve this answer
Ahh ok, now it makes more sense. After the transposition of the vector I can use it as a set, as far as I know. – beyeran Apr 22 '11 at 11:23

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.