Transcript Chapter 8
Chapter 8
Estimation
©
Estimator and Estimate
An estimator of a population parameter is a
random variable that depends on the sample
information and whose value provides
approximations to this unknown parameter.
A specific value of that random variable is
called an estimate.
Point Estimator and Point
Estimate
Let represent a population parameter (such as
the population mean or the population
proportion ). A point estimator, θ̂ , of a
population parameter, , is a function of the
sample information that yields a single number
called a point estimate. For example, the sample
mean, X, is a point estimator of the population
mean , and the value that X assumes for a given
set of data is called the point estimate.
Unbiasedness
The point estimator θ̂ is said to be an unbiased
estimator of the parameter if the expected
value, or mean, of the sampling distribution of θ̂
is ; that is,
E (ˆ)
Probability Density Functions for
unbiased and Biased Estimators
(Figure 8.1)
ˆ1
ˆ2
ˆ
Bias
Let θ̂ be an estimator of . The bias in θ̂ is defined as
the difference between its mean and ; that is
Bias (ˆ) E (ˆ)
It follows that the bias of an unbiased estimator is 0.
Most Efficient Estimator and
Relative Efficiency
Suppose there are several unbiased estimators of . Then the
unbiased estimator with the smallest variance is said to
be the most efficient estimator or to be the minimum
variance unbiased estimator of . Let θ̂1and θ̂ 2 be two
unbiased estimators of , based on the same number of
sample observations. Then,
a) θ̂1is said to be more efficient than θ̂ 2 if Var (ˆ1 ) Var (ˆ2 )
b) The relative efficiency of θ̂1 with respect to θ̂ 2 is the ratio of
their variances; that is,
Var(θˆ2 )
Relative Efficiency
Var(θˆ1 )
Point Estimators of Selected
Population Parameters
(Table 8.1)
Population
Parameter
Point
Estimator
Properties
Mean,
X
Unbiased, Most Efficient
(assuming normality)
Mean,
Xm
Unbiased (assuming
normality), but not most
efficient
Proportion,
p
Unbiased, Most Efficient
Variance, 2
s2
Unbiased, Most Efficient
(assuming normality)
Confidence Interval Estimator
A confidence interval estimator for a population
parameter is a rule for determining (based on
sample information) a range, or interval that is
likely to include the parameter. The corresponding
estimate is called a confidence interval estimate.
Confidence Interval and
Confidence Level
Let be an unknown parameter. Suppose that on the basis
of sample information, random variables A and B are found
such that P(A < < B) = 1 - , where is any number
between 0 and 1. If specific sample values of A and B are a
and b, then the interval from a to b is called a 100(1 - )%
confidence interval of . The quantity of (1 - ) is called the
confidence level of the interval.
If the population were repeatedly sampled a very
large number of times, the true value of the parameter
would be contained in 100(1 - )% of intervals calculated
this way. The confidence interval calculated in this manner
is written as a < < b with 100(1 - )% confidence.
P(-1.96 < Z < 1.96) = 0.95, where Z is
a Standard Normal Variable
(Figure 8.3)
0.95 = P(-1.96 < Z < 1.96)
0.025
0.025
-1.96
1.96
Notation
Let Z/2 be the number for which
P ( Z Z / 2 )
2
where the random variable Z follows a standard
normal distribution.
Selected Values Z/2 from the
Standard Normal Distribution Table
(Table 8.2)
Z/2
Confidence
Level
0.01
0.02
0.05
0.10
2.58
2.33
1.96
1.645
99%
98%
95%
90%
Confidence Intervals for the Mean of a
Population that is Normally Distributed:
Population Variance Known
Consider a random sample of n observations from a normal
distribution with mean and variance 2. If the sample
mean is X, then a 100(1 - )% confidence interval for the
population mean with known variance is given by
or equivalently,
Z / 2
Z / 2
X
X
n
n
X B
where the margin of error (also called the sampling error,
the bound, or the interval half width) is given by
B Z / 2
n
Basic Terminology for Confidence
Interval for a Population Mean with
Known Population Variance
(Table 8.3)
Terms
Symbol
Standard Error of the Mean
X
Z Value (also called the
Reliability Factor)
Z / 2
Margin of Error
Lower Confidence Limit
Upper Confidence Limit
Width (width is twice the bound)
B
To Obtain:
/ n
Use Standard Normal
Distribution Table
B Z / 2
n
LCL
LCL X Z / 2
UCL
UCL X Z / 2
w
w 2 B 2 Z / 2
n
n
n
Student’s t Distribution
Given a random sample of n observations, with
mean X and standard deviation s, from a normally
distributed population with mean , the variable t
follows the Student’s t distribution with (n - 1)
degrees of freedom and is given by
X
t
s/ n
Notation
A random variable having the Student’s t
distribution with v degrees of freedom will be
denoted tv. The tv,/2 is defined as the number
for which
P (tv tv , / 2 ) / 2
Confidence Intervals for the Mean of a
Normal Population: Population Variance
Unknown
Suppose there is a random sample of n observations from a normal
distribution with mean and unknown variance. If the sample
mean and standard deviation are, respectively, X and s, then a 100(1
- )% confidence interval for the population mean, variance
unknown, is given by
X tn 1, / 2
s
s
X tn 1, / 2
n
n
or equivalently,
X B
where the margin of error, the sampling error, or bound, B, is given
s
by
B t n 1, / 2
n
and tn-1,/2 is the number for which
P (t n 1 t n 1, / 2 ) / 2
The random variable tn-1 has a Student’s t distribution with v=(n-1) degrees of freedom.
Confidence Intervals for Population
Proportion (Large Samples)
Let p denote the observed proportion of “successes” in a random
sample of n observations from a population with a proportion of
successes. Then, if n is large enough that (n)()(1- )>9, then a 100(1
- )% confidence interval for the population proportion is given
by
p Z / 2
p(1 p)
p(1 p)
p Z / 2
n
n
or equivalently,
pB
where the margin of error, the sampling error, or bound, B, is given
by
p(1 p)
B Z / 2
n
and Z/2, is the number for which a standard normal variable Z
satisfies
P ( Z Z / 2 ) / 2
Notation
A random variable having the chi-square
distribution with v = n-1 degrees of freedom
will be denoted by 2v or simply 2n-1. Define
as 2n-1, the number for which
P(
2
n 1
2
n 1,
)
The Chi-Square Distribution
(Figure 8.17)
1-
0
2n-1,
The Chi-Square Distribution for n – 1
and (1-)% Confidence Level
(Figure 8.18)
/2
/2
1-
2n-1,1- /2
2n-1,/2
Confidence Intervals for the Variance of a
Normal Population
Suppose there is a random sample of n observations from a
normally distributed population with variance 2. If the observed
variance is s2 , then a 100(1 - )% confidence interval for the
population variance is given by
(n 1) s 2
n21, / 2
2
(n 1) s 2
n21,1 / 2
is the number for which
P(
and 2n-1,1 - /2 is the number for which
P(
where
2
n-1,/2
2
n 1
2
n 1
2
n 1, / 2
)
2
n 1,1 / 2
2
)
And the random variable 2n-1 follows a chi-square distribution
with (n – 1) degrees of freedom.
2
Confidence Intervals for Two Means:
Matched Pairs
Suppose that there is a random sample of n matched pairs of
observations from a normal distributions with means X and Y .
That is, x1, x2, . . ., xn denotes the values of the observations from the
population with mean X ; and y1, y2, . . ., yn the matched sampled
values from the population with mean Y . Let d and sd denote the
observed sample mean and standard deviation for the n differences
di = xi – yi . If the population distribution of the differences is
assumed to be normal, then a 100(1 - )% confidence interval for
the difference between means (d = X - Y) is given by
d tn 1, / 2
or equivalently,
sd
sd
d d tn 1, / 2
n
n
d B
Confidence Intervals for Two Means:
Matched Pairs
(continued)
Where the margin of error, the sampling error or the bound,
B, is given by
B t n 1, / 2
sd
n
And tn-1,/2 is the number for which
P (t n 1 t n 1, / 2 )
2
The random variable tn – 1, has a Student’s t distribution
with (n – 1) degrees of freedom.
Confidence Intervals for Difference Between
Means: Independent Samples (Normal
Distributions and Known Population Variances)
Suppose that there are two independent random samples of nx and
ny observations from normally distributed populations with means
X and Y and variances 2x and 2y . If the observed sample means
are X and Y, then a 100(1 - )% confidence interval for (X - Y) is
given by
( X Y ) Z / 2
or equivalently,
X2
nx
Y2
ny
X Y ( X Y ) Z / 2
(X Y ) B
where the margin of error is given by
B Z / 2
X2
nx
Y2
ny
X2
nx
Y2
ny
Confidence Intervals for Two Means:
Unknown Population Variances that are
Assumed to be Equal
Suppose that there are two independent random samples with nx and
ny observations from normally distributed populations with means X
and Y and a common, but unknown population variance. If the
observed sample means are X and Y, and the observed sample
variances are s2X and s2Y, then a 100(1 - )% confidence interval for (X
- Y) is given by
s 2p s 2p
s 2p s 2p
( X Y ) t nx n y 2, / 2
X Y ( X Y ) t nx n y 2, / 2
nx n y
nx n y
or equivalently,
(X Y ) B
where the margin of error is given by
B t nx n y 2, / 2
s 2p
nx
s 2p
ny
Confidence Intervals for Two Means: Unknown
Population Variances that are Assumed to be Equal
(continued)
The pooled sample variance, s2p, is given by
s
2
p
tnx ny 2, / 2 is the number for which
(nx 1) s X2 (n y 1) sY2
nx n y 2
P(t nx n y 2 t nx n y 2, / 2 )
2
The random variable, T, is approximately a Student’s t distribution
with nX + nY –2 degrees of freedom and T is given by,
( X Y ) ( X Y )
T
1
1
sp
n X nY
Confidence Intervals for Two Means:
Unknown Population Variances, Assumed
Not Equal
Suppose that there are two independent random samples of nx and ny
observations from normally distributed populations with means X
and Y and it is assumed that the population variances are not equal.
If the observed sample means and variances are X, Y, and s2X , s2Y, then
a 100(1 - )% confidence interval for (X - Y) is given by
( X Y ) t( v , / 2 )
s X2 sY2
s X2 sY2
X Y ( X Y ) t( v , / 2 )
nx n y
nx n y
where the margin of error is given by
B t( v , / 2 )
s X2 sY2
nx n y
Confidence Intervals for Two Means: Unknown
Population Variances, Assumed Not Equal
(continued)
The degrees of freedom, v, is given by
s X2
sY2 2
[( ) ( )]
nX
nY
v 2
sX 2
sY2 2
( ) /( n X 1) ( ) /( nY 1)
nX
nY
If the sample sizes are equal, then the degrees of freedom reduces to
2
(n 1)
v 1 2
s X sY2
2
2
sY s X
Confidence Intervals for the Difference
Between Two Population Proportions
(Large Samples)
Let pX, denote the observed proportion of successes in a random
sample of nX observations from a population with proportion X
successes, and let pY denote the proportion of successes observed in
an independent random sample from a population with proportion
Y successes. Then, if the sample sizes are large (generally at least
forty observations in each sample), a 100(1 - )% confidence interval
for the difference between population proportions (X - Y) is given
by
( pX pY ) B
Where the margin of error is
B Z / 2
p X (1 p X ) pY (1 pY )
nX
nY
Sample Size for the Mean of a
Normally Distributed Population
with Known Population Variance
Suppose that a random sample from a normally
distributed population with known variance 2 is
selected. Then a 100(1 - )% confidence interval for
the population mean extends a distance B
(sometimes called the bound, sampling error, or the
margin of error) on each side of the sample mean, if
the sample size, n, is
Z / 2
n
2
B
2
2
Sample Size for Population
Proportion
Suppose that a random sample is selected from a
population. Then a 100(1 - )% confidence interval
for the population proportion, extending a distance
of at most B on each side of the sample proportion,
can be guaranteed if the sample size, n, is
0.25( Z / 2 )
n
B2
2
Key Words
Bias
Bound
Confidence interval:
For mean, known variance
For mean, unknown
variance
For proportion
For two means, matched
For two means, variances
equal
For two means, variances
not equal
For variance
Confidence Level
Estimate
Estimator
Interval Half Width
Lower Confidence Limit
(LCL)
Margin of Error
Minimum Variance
Unbiased Estimator
Most Efficient Estimator
Point Estimate
Point Estimator
Key Words
(continued)
Relative Efficiency
Reliability Factor
Sample Size for Mean,
Known Variance
Sample Size for
Proportion
Sampling Error
Student’s t
Unbiased Estimator
Upper Confidence Limit
(UCL)
Width