9
$\begingroup$

I have always heard "orthogonal" in the area of geometry (also please note I am not a native English speaker). I do not understand the following for latin squares (a quote from a text book):

Every treatment (ABCD) appears once in each row. Hence the treatments and rows are orthogonal. ... Rows and columns are orthogonal to treatments.

\begin{matrix}\,&1&2&3&4\\1&A&B&C&D\\2&B&C&D&A\\3&C&D&A&B\\4&D&A&B&C\end{matrix}

What is meant by orthogonality here?

$\endgroup$
3
  • 2
    $\begingroup$ Possible duplicate: stats.stackexchange.com/questions/228797/… $\endgroup$
    – Glen_b
    Commented Jun 20, 2017 at 15:40
  • 2
    $\begingroup$ This question relates specifically to latin squares, the "duplicate" asks about orthogonality in general. I would think the upvotes and missing answer indicate that is not answered by the one you referenced. $\endgroup$
    – John V
    Commented Jun 30, 2017 at 16:38
  • 2
    $\begingroup$ See answer in stats.stackexchange.com/questions/286675/… $\endgroup$
    – F. Tusell
    Commented Jul 4, 2017 at 17:53

1 Answer 1

2
$\begingroup$

what it means, or, what the latin square does

The orthogonality of columns $i$ and rows $j$ means that their effect is being removed from the expectation values for some treatment $k$ (A,B,C,D).

See the formula (for a model without cross effects)

$Y_{ijk} = \alpha + c_i + r_j + \beta_k + \epsilon_{ijk}$

whose expectation for a certain level of $k$ (A, B, C or D) becomes the following

$E(Y_{ijk} \vert k) = \alpha + \beta_k$

provided that the treatment does not correlate (is orthogonal to) with the rows and columns.

the treatment for A (and similarly for B, C and D) is tested the same number of times in each row and so you can eliminate (average out) the effect of the row on the expectation value of the treatment A.

orthogonality

I am not sure if this is the origin of the etymology but this is what I imagine with orthogonality

In the example you have the following tests (column, row, treatment):

1,1,A
1,2,B
1,3,C
1,4,D
2,1,B
2,2,C
2,3,D
2,4,A
3,1,C
3,2,D
3,3,B
3,4,A
4,1,D
4,2,A
4,3,B
4,4,C

if you take this as a matrix $M$ and calculate $M^TM$ then you obtain in the non-diagonal elements a sum of products in which each term occurs the same number of times.

for example the product of the first and third column $(1,1,1,1,2,2,2,2,3,3,3,3,4,4,4,4) \cdot (A,B,C,D,B,C,D,A,C,D,A,B,D,A,B,C) = (1+2+3+4)(A+B+C+D) = 16 \mu_i \mu_j $

and this property may be associated with orthogonality of columns in a matrix

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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