8
$\begingroup$

I am struggling to fully understand some notation in a book where they use a "crosshair" symbol - first like $\bigoplus\limits_{i=1}^n{} Z_j $ where the $Z_j$ are matrices and second like $I_n \otimes \Phi$ where $I_n$ and $\Phi$ are both matrices.

The book is about multivariate statistics and the section is about random coefficient models. There isn't a notation/terminology appendix to refer to. I was going to post a digital pic of the page so that users can see the context (this is at the start of the section).

So is this on topic here or should I post on math.se ?

Update: I originally posted this on meta.se and it was migrated here. I am now attaching the photo from the relevant page of the book. enter image description here

Improve this question
$\endgroup$
4
  • 8
    $\begingroup$ $\oplus$, Direct sum; $\otimes$, Tensor product. $\endgroup$
    – whuber ♦
    Oct 14, 2014 at 15:41
  • $\begingroup$ @whuber thanks ! That's exactly what I needed. Did you want to make it into an answer so I can accept it ? $\endgroup$
    – Joe King
    Oct 14, 2014 at 15:53
  • $\begingroup$ I'm too busy now, Joe--a real answer would explain what these things are, rather than resorting to links. If anyone would care to provide the details in an answer I'd be happy to vote it up, but in the meantime I am glad you can move on with your reading. $\endgroup$
    – whuber ♦
    Oct 14, 2014 at 15:54
  • $\begingroup$ @whuber OK I will upload the digital picture I mentioned to give more context $\endgroup$
    – Joe King
    Oct 14, 2014 at 16:14

1 Answer 1

Reset to default
10
$\begingroup$

In statistics, $$A\oplus B:=\left[\begin{array}{cc}A & 0 \\ 0 & B \end{array} \right]$$ and (e.g. for a $2\times 2$-matrix $A$) $$A\otimes B:=\left[\begin{array}{cc}a_{11}B & a_{12}B \\ a_{21}B & a_{22}B \end{array} \right].$$

This focuses on matrices for their use in statistics as design or hypothesis matrices etc., where these notations simplify the frequent block structure of such matrices. One can find the name Kronecker sum for $\oplus$ and Kronecker product for $\otimes$, especially in the manuals of statistical software. (Also very handy is the component wise matrix multiplication $A\#B=[a_{ij}b_{ij}]_{i,j}$ for equally shaped matrices. It's sometimes called Hadamard product.)

In mathematics, $\oplus$ and $\otimes$ have their slightly different typical meaning as direct sum or tensor product of vector spaces or even more general algebraic structures.

Improve this answer
$\endgroup$
6
  • 3
    $\begingroup$ The statistical meaning is identical to the mathematical meaning of direct sum and tensor product of linear transformations: these formulas result from writing out the transformations as matrices in a particular basis. $\endgroup$
    – whuber ♦
    Oct 15, 2014 at 18:46
  • $\begingroup$ The theorem holds that the set of $n\times m$-matrices is isomorphic to the set of linear transformations from a $m$ dimensional to a $n$ dimensional vector space. So in the vector space world you are right. But a direct sum exists also between structures like groups. In statistics, the latter is usually not meant. $\endgroup$
    – Horst Grünbusch
    Oct 16, 2014 at 8:18
  • $\begingroup$ @whuber when I first searched for this I came upon [this] (math.stackexchange.com/questions/207635/…) which confused me quite a bit and prompted me to ask the question on CV. Is that link actually showing the same usage, because the accepted anser there says it is "non carrying addition" ? $\endgroup$
    – Joe King
    Oct 16, 2014 at 8:19
  • $\begingroup$ @HorstGrünbusch I just wanted to alert you to the comment I just made above toe whuber - also, if you could update your answer to reflect what the meanings are called in statistics I would be happy to accept it. $\endgroup$
    – Joe King
    Oct 16, 2014 at 8:23
  • $\begingroup$ I never came across this particular meaning of $\oplus$ and I'm not even sure if the accepted answer there is right. But I'm quite sure it is not the Kronecker sum from matrix arithmetic. $\endgroup$
    – Horst Grünbusch
    Oct 16, 2014 at 8:35

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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