This action might not be possible to undo. Are you sure you want to continue?

BooksAudiobooksComicsSheet Music### Categories

### Categories

### Categories

Editors' Picks Books

Hand-picked favorites from

our editors

our editors

Editors' Picks Audiobooks

Hand-picked favorites from

our editors

our editors

Editors' Picks Comics

Hand-picked favorites from

our editors

our editors

Editors' Picks Sheet Music

Hand-picked favorites from

our editors

our editors

Top Books

What's trending, bestsellers,

award-winners & more

award-winners & more

Top Audiobooks

What's trending, bestsellers,

award-winners & more

award-winners & more

Top Comics

What's trending, bestsellers,

award-winners & more

award-winners & more

Top Sheet Music

What's trending, bestsellers,

award-winners & more

award-winners & more

Welcome to Scribd! Start your free trial and access books, documents and more.Find out more

**STA705 Spring 2006
**

Let f(x) be a function (possibly multivariate) and suppose we are interested in determining

the maximum of f and, often more importantly, the value of x which maximizes f. The most

common statistical application of this problem is ﬁnding a Maximum Likelihood Estimate (MLE).

This document discusses the Newton Raphson method

1 Motivation

Newton Raphson maximization is based on a Taylor series expansion of the function f(x). Specif-

ically, if we expand f(x) about a point a,

f(x) ≈ f(a) + (x −a)

T

f

(a) +

1

2

(x −a)

T

f

(a)(x −a)

where f

(·) is the gradient vector and f

**(·) is the hessian matrix of second derivatives. This creates
**

a quadratic approximation for f. We know how to maximize a quadratic function (take derivatives,

set equal to zero, and solve)

d

dx

f(a) + (x −a)

T

f

(a) +

1

2

(x −a)

T

f

(a)(x −a) = f

(a) + (x −a)

T

f

(a) = 0

x = a −[f

(a)]

T

[f

(a)]

−1

The Newton Raphson process iterates this equation. Speciﬁcally, let x

0

be a starting point for

the algorithm and deﬁne successive estimates x

1

, x

2

, . . . recursively through the equation

x

i+1

= x

i

−[f

(x

i

)]

T

[f

(x

i

)]

−1

If the function f(x) is quadratic, then of course the quadratic “approximation” is exact and the

Newton Raphson method converges to the maximum in one iteration. If the function is concave,

then the Newton Raphson method is gauranteed to converge to the correct answer. If the function

is convex for some values of x, then the algorithm may or may not converge. The NR algorithm may

converge to a local maximum and not the global maximum, it might converge to a local minimum,

or it might cycle between two points. Starting the algorithm near the global maximum is the best

practical method for helping convergence to the global maximum.

Fortunately, loglikelihoods are typically approximately quadratic (the reason asymptotically

normality occurs for many random variables). Thus, the NR algorithm is an obvious choice for

ﬁnding MLEs. The starting value for the algorithm is often a “simpler” estimate (in terms of ease

of computation) of the parameter, such as a method of moments estimator.

1

Example

Let X

1

, . . . , X

n

∼ Gamma(α, β) and suppose we want the joint maximum likelihood estimate

of (α, β). The loglikelihood is

lnf(x|α, β) = ln

_

β

nα

Γ

n

(α)

_

x

i

_

α−1

exp

_

−β

x

i

_

_

= nαlnβ −nlnΓ(α) + (α −1)

lnx

i

−β

x

i

Solving for the MLE analytically requires solving the equations

∂

∂α

lnf(x|α, β) = nlnβ −n

∂ lnΓ(α)

∂α

+

lnx

i

= 0

∂

∂β

lnf(x|α, β) =

nα

β

−

x

i

= 0

These two equations cannot be solved analytically (the Gamma function is diﬃcult to work with).

The two equations do provide us with the gradient

f

(α, β) =

_

nlnβ −n

∂ ln Γ(α)

∂α

+

lnx

i

(nα/β) −

x

i

_

The hessian matrix is

f

(α, β) =

_

−n

∂

2

ln Γ(α)

∂α

2

n/β

n/β −nα/β

2

_

The starting values of the algorithm may be found using the method of moments. Since E[X

i

] =

α/β and V [X

i

] = α/β

2

, the method of moment estimators are α

M

= ¯ x

2

/s

2

and β

M

= ¯ x/s

2

.

Suppose we have data (in truth actually generated with α = 2 and β = 3) such that n = 1000,

ln

X

i

= −646.0951,

¯

X = 0.6809364, and s

2

= 0.2235679. The algorithm begins at α

M

=

2.073976 and β

M

= 3.04577. After 3 iterations, the NR algorithm stabilizes at ˆ α = 2.060933 and

ˆ

β = 3.026616.

Note that there is no need to treat this situation as a multivariate problem. The partial

derivative for β may be solved in terms of α.

∂

∂β

lnf(x|α, β) =

nα

β

−

x

i

= 0

β = nα/

x

i

= α/¯ x

Thus, for each value of α, the maximum of β is attained at β = nα/

x

i

. Thus, we can reduce the

problem to the one-dimensional problem of maximizing

2

lnf(x|α, β = α/¯ x) = nαln(α/¯ x) −nlnΓ(α) + (α −1)

lnx

i

−(α/¯ x)

x

i

= nαlnα −nαln ¯ x −nlnΓ(α) + (α −1)

lnx

i

−nα

This is called the proﬁle loglikelihood. The ﬁrst and second derivatives are

∂

∂α

= n[1 + lnα] −nln ¯ x −n

∂ lnΓ(α)

∂α

+

lnx

i

−n = nlnα −nln ¯ x −n

∂ lnΓ(α)

∂α

+

lnx

i

∂

2

∂α

2

=

n

α

−n

∂

2

lnΓ(α)

∂α

2

Again starting the algorithm at α

M

= 2.073976, the algorithm converges to ˆ α = 2.060933 after

???? iterations.

3

Newton Raphson Maximum Likelihoo

Newton Raphson Maximum Likelihoo

Are you sure?

This action might not be possible to undo. Are you sure you want to continue?

We've moved you to where you read on your other device.

Get the full title to continue

Get the full title to continue listening from where you left off, or restart the preview.

scribd