Embed
Email

Matrix Algebra

Document Sample
Matrix Algebra
Shared by: HC111214034914
Categories
Tags
Stats
views:
0
posted:
12/13/2011
language:
pages:
46
Matrix Algebra

 Matrix algebra is a means of expressing large

numbers of calculations made upon ordered

sets of numbers.

 Often referred to as Linear Algebra

 Many equations would be completely

intractable if scalar mathematics had to be

used. It is also important to note that the

scalar algebra is under there somewhere.

Definitions - scalar

 scalar - a number

 denoted with regular type as is scalar

algebra

 [1] or [a]

Definitions - vector

 vector - a single row or column of numbers

 denoted with bold small letters

 row vector a = 1 2 3 4 5

 column vector x =

 x1 

x 

 2

 x3 

 

 x4 

 x5 

 

Definitions - Matrix

 A matrix is a set of rows and columns of

numbers 1 2 3

 4 5 6

 

 Denoted with a bold Capital letter

 All matrices (and vectors) have an order

- that is the number of rows x the

number of columns.

 Thus A = 1 2 3

 4 5 6

  2 x 3 

Matrix Equality

 Two matrices are equal iff (if and only

if) all of their elements are identical

 Note: your data set is a matrix.

Matrix Operations

 Addition and Subtraction

 Multiplication

 Transposition

 Inversion

Addition and Subtraction

 Two matrices may be added iff they are

the same order.

 Simply add the corresponding elements



a11 a12   b11 b12  c11 c12 

a a22   b21 b22   c 21 c 22 

 21     

a31 a32  b31 b32  c31 c32 

     

Addition and Subtraction

(cont.)

a11  b11  c11

 Where a12  b12  c12

a21  b21  c 21

a22  b22  c 22

a31  b31  c31

a32  b32  c32







 Hence 1 2 4 6 5 8 

3 4  4 6  7 10

     

5 6 4 6 9 12

     

Scalar Multiplication

 To multiply a scalar times a matrix,

simply multiply each element of the

matrix by the scalar quantity



a11 a12   2a11 2a12 

2   2a

a21 a22   21 2a22 

 

Matrix Multiplication (cont.)

 To multiply a matrix times a matrix, we

write

 A times B as AB

 This is pre-multiplying B by A, or post-

multiplying A by B.

Matrix Multiplication (cont.)

 In order to multiply matrices, they must

be conformable (the number of columns

in A must equal the number of rows in

B.)

 an (mxn) x (nxp) = (mxp)

 an (mxn) x (pxn) = cannot be done

 a (1xn) x (nx1) = a scalar (1x1)

Matrix Multiplication (cont.)

 The general rule for

matrix multiplication

is:

N

cij   aik bkj where i  1,2,..., M , and j  1,2,..., P

k 1

Matrix multiplication is not

Commutative

 AB does not necessarily equal BA

 (BA may even be an impossible

operation)

Yet matrix multiplication is

Associative

 A(BC) = (AB)C

Special matrices

 There are a number of special matrices

 Square

 Diagonal

 Symmetric

 Null

 Identity

Square matrix

 A square matrix is just what it sounds like, an nxn

matrix

 a11 a12 a13 a14 

a 

 21 a22 a23 a24 

a31 a32 a33 a34 

 

a41 a42 a43 a44 

 Square matrices are quite useful for describing the

properties or interrelationships among a set of

things – like a data set.

Diagonal Matrices

 A diagonal matrix is a square matrix that

has values on the diagonal with all off-

diagonal entities being zero.



a11 0 0 0

0 a 0 0

 22 

0 0 a33 0

 

0 0 0 a44 

Symmetric Matrix

 All of the elements in the upper right portion

of the matrix are identical to those in the

lower left.

 For example, the correlation matrix

Identity Matrix

 The identity matrix I is a diagonal

matrix where the diagonal elements all

equal one. It is used in a fashion

analogous to multiplying through by "1"

in scalar math.

1 0 0 0

0 1 0 0

 

0 0 1 0

 

0 0 0 1

Null Matrix

 A square matrix where all elements equal

zero.

0 0 0 0

0 0 0 0

 

0 0 0 0

 

0 0 0 0

 Not usually „used‟ so much as sometimes the

result of a calculation.

 Analogous to “a+b=0”

The Transpose of a Matrix A'

 Taking the transpose is an operation

that creates a new matrix based on an

existing one.

 The rows of A = the columns of A'

 Hold upper left and lower right corners

and rotate 180 degrees.

Example of a transpose





1 4 

2 5, A'  1 2 3

A   4 5 6

3 6  

 

The Transpose of a Matrix A'

 If A = A', then A is symmetric (i.e.

correlation matrix)

 If AA‟ = A then A' is idempotent

 (and A' = A)

 ( A  B C )' sum B C'

The transposeof a  A'= 'sum of transposes

 The transpose of a product = the product of

the transposes in reverse order

 (ABC)‟ = C‟B‟A‟

An example:

 Suppose that you wish to obtain the

sum of squared errors from the vector

e. Simply pre-multiply e by its

transpose e'.

 which, in matrices looks like





e' e  e1  e2  ..en

2 2 2



An example - cont

 Since the matrix product is a scalar

found by summing the elements of the

vector squared.

The Determinant of a Matrix

 The determinant of a matrix A is

denoted by |A|.

 Determinants exist only for square

matrices.

 They are a matrix characteristic, and

they are also difficult to compute

The Determinant for a 2x2 matrix

 If A = a11 a12 

a 

 21 a22 









 Then

A  a11a22  a12a21





 That one is easy

The Determinant for a 3x3 matrix

 If A =

a11 a12 a13 

a 

 21 a22 a23 

a31 a32

 a33 





 Then



A  a11a22a33  a11a23a32  a12a23a31  a12a21a33  a13a21a32  a13a22a31

Determinants

 For 4 x 4 and up don't try. For those

interested, expansion by minors and

cofactors is the preferred method.

 (However the spaghetti method works

well! Simply duplicate all but the last

column of the matrix next to the

original and sum the products of the

diagonals along the following pattern.)

Spaghetti Method of |A|





a11 a12 a13  a11 a12 

a a22  a 

a23   21 a22 

 21



a31 a32

 a33  a31 a32 

 

Properties of Determinates

 Determinants have several mathematical

properties which are useful in matrix

manipulations.

 1 |A|=|A'|.

 2. If a row of A = 0, then |A|= 0.

 3. If every value in a row is multiplied by k,

then |A| = k|A|.

 4. If two rows (or columns) are interchanged

the sign, but not value, of |A| changes.

 5. If two rows are identical, |A| = 0.

Properties of Determinates

 6. |A| remains unchanged if each

element of a row or each element

multiplied by a constant, is added to any

other row.

 7. Det of product = product of Det's

|AB| = |A| |B|

 8. Det of a diagonal matrix = product

of the diagonal elements

The Inverse of a Matrix (A-1)

 For an nxn matrix A, there may be a B such

that AB = I = BA.

 The inverse is analogous to a reciprocal)

 A matrix which has an inverse is nonsingular.

 A matrix which does not have an inverse is

singular.

 An inverse exists only if A  0

Inverse by Row or column

operations

 Set up a tableau matrix

 A tableau for inversions consists of the

matrix to be inverted post multiplied by

a conformable identity matrix.

Matrix Inversion by Tableau

Method

 Rules:

 You may interchange rows.

 You may multiply a row by a scalar.

 You may replace a row with the sum of that row

and another row multiplied by a scalar.

 Every operation performed on A must be

performed on I

 When you are done; A = I & I = A-1

The Tableau Method of Matrix

Inversion: An Example

 Step 1: Set up Tableau 1 4 3  1 0 0

2 5 4  0 1 0 

  

1  3  2  0 0 1 

  





 Step 2: Add –2(Row 1) 1 4 3   1 0 0

0  3  2    2 1 0 

to Row 2   

1  3  2  0 0 1

  

Matrix Inversion – cont.

1 4 3   1 0 0

0  3  2    2 1 0 

  

1  3  2  0 0 1

  



 Step 3: Add –1(Row 1 4 3   1 0 0

0  3  2    2 1 0 

1) to Row 3   

0  7  5    1 0 1 

  

 Step 4: Multiply Row 1 4 3  1 0 0

0 1 2 / 3 2 / 3  1 / 3 0

2 by –1/3   

0  7  5    1

  0 1



Matrix Inversion – cont.

1 4 3  1 0 0

0 1 2 / 3 2 / 3  1 / 3 0

  

0  7  5    1

  0 1



 Step 5: Add –4 (Row 2) 1 0 1 / 3    5 / 3 4 / 3 0

to Row 1 0 1 2 / 3  2 / 3  1 / 3 0

  

0  7  5    1

  0 1



1 0 1 / 3    5 / 3 4 / 3 0

 Step 6: Add 7(Row 2) 0 1 2 / 3   2 / 3  1 / 3 0

  

to Row 3 0 0  1 / 3  11 / 3  7 / 3 1

  

Matrix Inversion – cont.

1 0 1 / 3    5 / 3 4 / 3 0

0 1 2 / 3   2 / 3  1 / 3 0

  

0 0  1 / 3  11 / 3  7 / 3 1

  



 Step 7: Add Row 3 1 0 0  2  1 1

0 1 2 / 3   2 / 3  1 / 3 0

to Row 1   

0 0  1 / 3 11 / 3  7 / 3 1

  

1 0 0  2  1 1

 Step 9: Add 2(Row 0 1 0  8  5 2

  

3) to Row 2 0 0  1 / 3 11 / 3  7 / 3 1

  

Matrix Inversion – cont.

1 0 0  2  1 1

0 1 0  8  5 2

  

0 0  1 / 3 11 / 3  7 / 3 1

  

 Step 9: Multiply Row 3

by -3 1 0 0   2 1 1 

0 1 0   8  5 2 

  

 The original matrix is 0 0 1  11 7  3

  

now an identity, and

the original identity has

been transformed to the

Inverse

Checking the calculation

 Remember AA-1=I

1 4 3  2  1 1  1 0 0

2 5 4   8  5 2   0 1 0 

    

1  3  2  11 7  3 0 0 1

    







 Thus

1 * 2  4 * 8  3 * 11  1

1 * 1  4 * 5  3 * 7  0

etc

The Matrix Model

 The multiple regression model may be

easily represented in matrix terms.

Y  XB  e



 Where the Y, X, B and e are all

matrices of data, coefficients, or

residuals

The Matrix Model (cont.)

 The matrices in Y  XB  e are

represented by



 Y1   X 11 X 12 ... X ik   B1   e1 

Y  X X ... X 2 k  B  e 

Y  

2

X  21 22  B  

2

e  

2



  ... ... ... ...     

       

 Yn   X n1 X n 2 ... X nk   Bk   en 



 Note that we postmultiply X by B since this

order makes them conformable.

The Assumptions of the Model

Scalar Version

 1. The ei's are normally distributed.

 2. E(ei) = 0

 3. E(ei2) = 2

 4. E(eiej) = 0 (ij)

 5. X's are nonstochastic with values fixed in repeated

samples and (Xik-Xbark)2/n is a finite nonzero

number.

 6. The number of observations is greater than the

number of coefficients estimated.

 7. No exact linear relationship exists between any of

the explanatory variables.

The Assumptions of the

Model: The Matrix Version

 These same assumptions expressed in

matrix format are:



 1. e  N(0,)

 2.  = 2I

 3. The elements of X are fixed in repeated

samples and (1/ n)X'X is nonsingular and

its elements are finite

Derivation of B's in matrix

notation

 Given the matrix algebra model we can

replicate the least squares normal

equations in matrix format.

 We need to minimize ee’ which is the sum

of squared errors

 Setting the derivative equal to 0 we get

Note that X‟X is called the sums-of-squares

and cross-products matrix.


Related docs
Other docs by HC111214034914
Foundations
Views: 0  |  Downloads: 0
Subject Area:
Views: 0  |  Downloads: 0
Circuit Theorems
Views: 0  |  Downloads: 0
Academic Program Inventory
Views: 0  |  Downloads: 0
sdaaii2005 level 10
Views: 1  |  Downloads: 0
By registering with docstoc.com you agree to our
privacy policy

You are almost ready to download!

You are almost ready to download!