This article needs additional citations for verification. Please help improve this article by adding citations to reliable sources. Unsourced material may be challenged and removed.Find sources: "Weighing matrix" – news · newspapers · books · scholar · JSTOR (December 2013) (Learn how and when to remove this template message)

In mathematics, a **weighing matrix** of order and weight is a matrix with entries from the set such that:

Where is the transpose of and is the identity matrix of order . The weight is also called the *degree* of the matrix. For convenience, a weighing matrix of order and weight is often denoted by .^{[3]}

Weighing matrices are so called because of their use in optimally measuring the individual weights of multiple objects. When the weighing device is a balance scale, the statistical variance of the measurement can be minimized by weighing multiple objects at once, including some objects in the opposite pan of the scale where they subtract from the measurement.^{[1]}^{[2]}

Some properties are immediate from the definition. If is a , then:

- The rows of are pairwise orthogonal. Similarly, the columns are pairwise orthogonal.
- Each row and each column of has exactly non-zero elements.
- , since the definition means that , where is the inverse of .
- where is the determinant of .

A weighing matrix is a generalization of Hadamard matrix, which does not allow zero entries.^{[3]} As two special cases, a is a Hadamard matrix^{[3]} and a is equivalent to a conference matrix.

See also: Design of experiments § Example |

Weighing matrices take their name from the problem of measuring the weight of multiple objects. If a measuring device has a statistical variance of , then measuring the weights of objects and subtracting the (equally imprecise) tare weight will result in a final measurement with a variance of .^{[4]} It is possible to increase the accuracy of the estimated weights by measuring different subsets of the objects, especially when using a balance scale where objects can be put on the opposite measuring pan where they subtract their weight from the measurement.

An order matrix can be used to represent the placement of objects—including the tare weight—in trials. Suppose the left pan of the balance scale adds to the measurement and the right pan subtracts from the measurement. Each element of this matrix will have:

Let be a column vector of the measurements of each of the trials, let be the errors to these measurements each independent and identically distributed with variance , and let be a column vector of the true weights of each of the objects. Then we have:

Assuming that is non-singular, we can use the method of least-squares to calculate an estimate of the true weights:

The variance of the estimated vector cannot be lower than , and will be minimum if and only if is a weighing matrix.^{[4]}^{[5]}

Weighing matrices appear in the engineering of spectrometers, image scanners,^{[6]} and optical multiplexing systems.^{[5]} The design of these instruments involve an optical mask and two detectors that measure the intensity of light. The mask can either transmit light to the first detector, absorb it, or reflect it toward the second detector. The measurement of the second detector is subtracted from the first, and so these three cases correspond to weighing matrix elements of 1, 0, and −1 respectively. As this is essentially the same measurement problem as in the previous section, the usefulness of weighing matrices also applies.^{[6]}

Note that when weighing matrices are displayed, the symbol is used to represent −1. Here are some examples:

This is a :

This is a :

This is a :

Another :

Which is circulant, i.e. each row is a cyclic shift of the previous row. Such a matrix is called a and is determined by its first row.
Circulant weighing matrices are of special interest since their algebraic structure makes them easier for classification. Indeed, we know that a circulant weighing matrix of order and weight must be of square weight. So, weights are permissible and weights have been completely classified.^{[7]}
Two special (and actually, extreme) cases of circulant weighing matrices are (A) circulant Hadamard matrices which are conjectured not to exist unless their order is less than 5. This conjecture, the circulant Hadamard conjecture first raised by Ryser, is known to be true for many orders but is still open. (B) of weight and minimal order exist if is a prime power and such a circulant weighing matrix can be obtained by signing the complement of a finite projective plane.
Since all for have been classified, the first open case is .
The first open case for a general weighing matrix (certainly not a circulant) is .

Two weighing matrices are considered to be equivalent if one can be obtained from the other by a series of permutations and negations of the rows and columns of the matrix. The classification of weighing matrices is complete for cases where as well as all cases where are also completed.^{[8]} However, very little has been done beyond this with exception to classifying circulant weighing matrices.^{[9]}^{[10]}

This section possibly contains synthesis of material which does not verifiably mention or relate to the main topic. Relevant discussion may be found on the talk page. (November 2019) (Learn how and when to remove this template message)

There are many open questions about weighing matrices. The main question about weighing matrices is their existence: for which values of and does there exist a ? A great deal about this is unknown. An equally important but often overlooked question about weighing matrices is their enumeration: for a given and , how many 's are there?

This question has two different meanings. Enumerating up to equivalence and enumerating different matrices with same *n*,*k* parameters. Some papers were published on the first question but none were published on the second important question.