Basic Definitions An mn matrix A is a rectangular array of real numbers with m rows and n columns.

(Rows are horizontal and columns are vertical.) The numbers m and n are the dimensions of A. The real numbers in the matrix are called its entries. The entry in row i and column j is called aij or Aij.

Example Following is a 45 matrix with the entry A23 highlighted. 0 A= 1 2 0 3 2 -3 1

1/3 -1 10 1/3 3 2 1 1 0 0 1 0

Top of Page Operations with Matrices Transpose The transpose, AT, of a matrix A is the matrix obtained from A by writing its rows as columns. If A is an mn matrix and B = AT, then B is the nm matrix with bij = aji. Sum, Difference If A and B have the same dimensions, then their sum, A+B, is obtained by adding corresponding entries. In symbols, (A+B)ij = Aij + Bij. If A and B have the same dimensions, then their difference, A - B, is obtained by subtracting corresponding entries. In symbols, (A-B)ij = Aij - Bij. Scalar Multiple If A is a matrix and c is a number (sometimes called a scalar in this context), then the scalar multiple, cA, is obtained by multiplying every entry in A by c. In symbols, (cA)ij = c(Aij). Product If A has dimensions mn and B has dimensions np, then the product AB is defined, and has dimensions mp. The entry (AB)ij is obtained by multiplying row i of A by column j of B, which is done by multiplying corresponding entries together and then adding the results. Examples Transpose 0 1/3 = 1 -1 2 10

1/3 -1 10 Sum & Scalar Multiple

1 + 2

2 =

1/ 3 1 Product

2/ 3 2

5/ 3 5

0 1/3

1 1

1 2/3

1 2

2/3 -2 = 5/3 1/3

Visit our Matrix Algebra Tool for on-line matrix algebra computations. Top of Page

Algebra of Matrices The nn identity matrix is the matrix I that has 1's down the main diagonal and 0's everywhere else. In symbols, Iij = 1 if i = j and 0 if i j. A zero matrix is one whose entries are all 0. The various matrix operations, addition, subtraction, scalar multiplication and matrix multiplication, have the following properties.

Examples Following is the 44 identity matrix. 1 0 0 0 I= 0 1 0 0 0 0 1 0 0 0 0 1 The following illustrates the failure of the commutative law for matrix multiplication.

A+(B+C) = (A+B)+C A+B = B+A A+O = O+A = A A+( - A) = O = ( A)+A c(A+B) = cA+cB (c+d)A = cA+dA A=A 0A = O A(BC) = (AB)C AI = IA = A A(B+C) = AB + AC (A+B)C = AC + BC OA = AO = O (A+B)T = AT + BT (cA)T = c(AT) (AB)T = BTAT

Additive associative law Additive commutative law Additive identity law Additive inverse law Distributive law Distributive law Scalar unit Scalar zero Multiplicative associative law Multiplicative identity law Distributive law Distributive law Multiplication by zero matrix Transpose of a sum Transpose of a scalar multiple Transpose of a matrix product A=

1/3 -1 2/3

B=

-1

2/3 -2

AB =

-2

-1/3 5/3 -1/3 2

BA =

-2/3 8/3 Top of Page

The one rule that is conspicuously absent from this list is commutativity of the matrix product. In general, matrix multiplication is not commutative: AB is not equal to BA in general. Matrix Form of a System of Linear Equations An important application of matrix multiplication is this: The system of linear equations a11x1 + a12x2 + a13x3 + . . . + a1nxn a21x1 + a22x2 + a23x3 + . . . + a2nxn .............. am1x1 + am2x2 + am3x3 + . . . + amnxn = bm can be rewritten as the matrix equation AX = B where a11 a12 a13 a21 a22 a23 . . . 1n . . . a2n 1 1 -1 3 1 -1 1 1 -2 A= 3 2 -1 Top of Page x y z = 4 6 4 9 . = b1 = b2 Example The system x + y - z = 4 3x + y - z = 6 x + y - 2z = 4 3x + 2y - z = 9 has matrix form

.... ... am1 am2 am3 . . . amn X = [x1, x2, x3, . . . , xn]T B = [b1, b2, x3, . . . , bm]T Example The system of equations 1 2 4 2 4 6 4 6 8 has solution x AX = B by multiplying both sides by A-1, which gives us X = A-1B. y z = 1 2 4 1 = -2 1 -2 = 1/2 1/2 . Top of Page Determining Whether a Matrix is Invertible In order to determine whether an nn matrix A is invertible or not, and to find A 1 if it does exist, write down the n(2n) matrix [A | I] (this is A with the nn identity matrix set next to it). Row reduce this matrix. If the reduced form is [I | B] (i.e., has the identity matrix in the left part), then A is invertible and B = A-1. If you cannot obtain I in the left part, then A is singular Examples The matrix 1 2 4 A= 2 4 6 4 6 8 is invertible. The matrix 1 2 4 B= 2 4 6 2 4 7 is not. Top of Page 2 4 6 -2 2 1/2 4 6 8 1 -1/2 0
1

and

Matrix Inverse If A is a square matrix, one that has the same number of rows and columns, it is sometimes possible to take a matrix equation such as AX = B and solve for X by "dividing by A." Precisely, a square matrix A may have an inverse, written A-1, with the property that AA = A A = I. If A has an inverse we say that A is invertible, otherwise we say that A is singular. When A is invertible we can solve the equation
-1 -1

x y z =

1 1 -1

1 1 -1 1 1 -1

Inverse of a 22 Matrix The 22 matrix a A= c d b

Example 1 2 3 4 =
1

1 = (1)(4) - (2)(3) -2 1 .

4 -3

-2 1

is invertible if ad - bc is nonzero and is singular if ad - bc = 0. The number ad - bc is called the determinant of the matrix. When the matrix is invertible its inverse is given by the formula A
1

3/2 -1/2 Top of Page

1 = ad - bc

d -c

-b a . Example Paper, Scissors, Rock Rock beats (crushes) scissors; scissors beat (cut) paper, and paper beats (wraps) rock. Each +1 entry indicates a win for the row player, -1 indicates a loss, and 0 indicates a tie. Do you want to play? Click on a row strategy... Column Strategy

Two-Person Zero Sum Game In a two-person zero sum game, each of the two players is given a choice between several prescribed strategies at each turn, and each player's loss is equal to the other player's gain. The payoff matrix of a two-person zero sum game has rows labeled by the row player's strategies and columns labeled by the column player's strategies. The ij entry of the matrix is the payoff that accrues to the row player in the event that the row player uses strategy i and the column player uses strategy j.

0 Row Strategy 1 1

1 0 1

1 1 0

Top of Page Mixed Strategy, Expected Value A player uses a pure strategy if he or she uses the same strategy at each round of the game. A mixed strategy is a method of playing a game where the rows or columns are played at random so that each is used a given fraction of the time. We represent a mixed (or pure) strategy for the row player by a row matrix (probability vector) S = [a b c . . . ] with the same number of entries as there are rows, where each entry represents the fraction of times the corresponding row is played (or the probability of using that strategy) and where a + b + . . . = 1. Example Here is a variant of "paper, scissors, rock in which "paper/paper" and "rock/rock" is no longer a draw.

2 1 1

1 0 1

1 1 2

Suppose the row player uses the mixed strategy S = [0.75 0 0.25]

A mixed strategy for the column player is represented by a similar column matrix T. For both row and column players, pure strategies is represented by vectors in with a single 1 and the rest zeros. The expected value of the game with payoff matrix P corresponding to the mixed strategies S and T is given by e = SPT The expected value of the game is the average payoff per round if each player uses the associated mixed strategies for a large number of rounds.

(play paper 75% of the time, scissors 0% of the time and rock 25% of the time) and thje column player plays scissors and rock each 50% of the time; 0 T= 0.5

0.5 . Then the expected value of the game is e = SPT 2 = [0.75 0 0.25] 1 1 = -0.125 Top of Page 1 0 1 1 1 2 0 0.5 0.5

Minimax Criterion, Fundamental Principles of Game Theory Minimax Criterion A player using the minimax criterion chooses a strategy that, among all possible strategies, minimizes the effect of the other player's best counter-strategy. That is, an optimal (best) strategy according to the minimax criterion is one that minimizes the maximum damage the opponent can cause. Finding the minimax strategy is called solving the game. In the texbook we show a graphical method for solving 22 games. For general games, one uses the simplex method (see the Chapter 4 Summary). However, one can frequently simplify a game and sometimes solve it by "reducing by dominance" and/or checking whether it is "strictly determined" (see below). Fundamental Principles of Game Theory When analyzing any game, we make the following assumptions about both players: 1. Each player makes the best possible move. 2. Each player knows that his or her opponent is also making the best possible move

Example Consider the following game. Column Strategy A 1 Row 2 Strategy 3 0 0 1 B C

1 1 0 2

2 3

If the row player follows Principle 1, (s)he should never play Strategy 1 since Strategy 2 gives better payoffs no matter what strategy the column player chooses. (The payoffs in Row 2 are all at least as high as the coresponding ones in Row 1.) Further, following Principle 2, the row player expects that the column player will never play Strategy A, since Strategy B gives better payoffs as far as the column player is concerned. (The payoffs in Column B are all at least as low as the coresponding ones in Column A.) Top of Page

Reducing by Dominance One pure strategy dominates another if all its payoffs are more advantageous to the player than the corresponding ones in the other strategy. In terms of the payoff matrix, we can say it this way: A. Row r in the payoff matrix dominates row s if each payoff in row r is the corresponding

Example Consider the above game once again. Column Strategy A Row 1 0 B C

1 1

payoff in row s. B. Column r in the payoff matrix dominates column s if each payoff in row r is the corresponding payoff in columns.

Strategy

2 3

0 1

2 3

Since the entries in Row 2 are the corresponding ones in Row 1, Row 2 dominates Row 1. Since the entries in Column B are the corresponding ones in Column A, Column B dominates Column A. Top of Page

Saddle Point, Strictly Determined Game A saddle point is a payoff that is simultaneously a row minimum and a column maximum. To locate saddle points, circle the row minima and box the column maxima. The saddle points are those entries that are both circled and boxed. A game is strictly determined if it has at least one saddle point. The following statements are true about strictly determined games. A. All saddle points in a game have the same payoff value. B. Choosing the row and column through any saddle point gives minimax strategies for both players. In other words, the game is solved via the use of these (pure) strategies. The value of a strictly determined game is the value of the saddle point entry. A fair game has value of zero, otherwise it is unfair or biased. Input-Output Economic Models

Example Row Player:In the above game, there are two saddle points, shown in color. A 1 2 3 0 0 1 B C 1 1 0 2

2 3

Since the saddle point entries are zero, this is a fair game. Our on-line game theory utility can be used to check any game (up to 5 5) for saddle points. Try it out. Top of Page

An input-output matrix for an economy gives, as its jth column, the amounts (in dollars or other appropriate currency) of outputs of each sector used as input by sector j (for one year or other appropriate period of time). It also gives the total production of each sector of the economy for a year (called the production vector when written as a column). The technology matrix is the matrix obtained by dividing each column by the total production of the corresponding sector. Its ijth entry, the ijth technology coefficient, gives the input from sector i necessary to produce one unit of output from sector j. A demand vector is a column vector giving the total demand from outside the economy for the products of each sector. If A is the technology matrix, X is the production vector, and D is the demand vector, then (I - A)X = D, or X = (I - A)-1D.

These same equations hold if D is a vector representing change in demand, and X is a vector representing change in production. The entries in a column of (I - A)-1 represent the change in production in each sector necessary to meet a unit change of demand in the sector corresponding to that column, taking into account all direct and indirect effects.