Estimation: point and interval estimator
Estimation
Estimation
Point and interval estimate
Estimation: point and interval estimator
Estimation: terminology
Estimator, estimate, and estimation:
An estimator of a population parameter is a random variable that
depends on the sample information; its value provides approximations of
this unknown parameter.
A specific value of that random variable is called an estimate.
The process to obtain the value of estimator is called estimation
Estimation: point and interval estimator
Estimation: Point Estimation
Population characteristics are summarized by parameters – the true
values are typically unknown. For example, the population mean is
denoted by µ.
An estimation rule can be specified for a parameter of interest. This
estimation rule is called a point estimator.
For example, a point estimator for the population mean µ is:
X̄ = ni=1 Xni ;
P
The value that X̄ assumes for a given set of data is called the point
estimate, x̄.
– A point estimate is a numeric outcome
– Different samples of data will have different numeric observations and,
therefore, will result in different point estimates of the population
parameter
– A point estimator X̄ is a random variable and has a probability
distribution called sampling distribution.
Estimation: point and interval estimator
Point Estimator: properties
1 Unbiasedness: A point estimator X̄ is said to be unbiased if
E(X̄ ) = µ
The amount of bias can be calculated as
Bias(X̄ ) = E(X̄ ) − µ
2 Minimum variance/efficiency: Let X̄1 and X̄2 be two unbiased
estimators for population mean µ, then X̄1 is said to be more efficient
estimator if
var(X̄1 ) < var(X̄2 )
If X̄ is an unbiased estimator of µ , and no other unbiased estimator
has smaller variance than X̄ , then X̄ is said to be the most efficient or
minimum variance unbiased estimator of µ .
Estimation: point and interval estimator
Interval estimator
A confidence interval estimator for a population parameter is a rule
for determining (based on sample information) an interval, say
A < µ < B that is likely to include the parameter. The corresponding
estimate is called a confidence interval estimate.
An interval estimators have been described as being “likely” or “very
likely” to include the true, but unknown, value of the population
parameter
If the population is repeatedly sampled and intervals are calculated in
this fashion, then in the long run 95% (or some other percentage) of
the intervals would contain the true value of the unknown parameter.
The interval from A to B is then said to be a 95% confidence interval
estimator for the population parameter µ.
Estimation: point and interval estimator
Confidence interval and confidence level
Let µ be an unknown parameter. Suppose that on the basis of sample
information, random variables A and B are found such that
Pr [A < µ < B] = 1 − α,
where α is any number between 0 and 1. If the specific sample values of A
and B are a and b, then the interval from a to b is called a 100(1 − α)%
confidence interval of µ. The quantity 100(1 − α)% is called the
confidence level of the interval.
If the population is repeatedly sampled a very large number of times, the
true value of the parameter µ will be covered by 100(1 − α)% of intervals
calculated this way. The confidence interval calculated in this manner is
written as a < µ < b, with 100(1 − α)% confidence.
Estimation: point and interval estimator
Confidence interval for Normally distributed mean:
population variance is known
Let x1 , x2 . . . xn be a random sample of n observations from a normally
distributed population with unknown mean µ and known variance σ 2 .
Suppose that we want a 100(1 − α)% confidence interval of the
population mean. We saw before that
x̄ − µ x̄ − µ
Z= = √ ∼ N(0, 1)
SE(x̄) σ/ n
Zα/2 is the value from the standard normal distribution such that the
upper tail probability is α/2.
Estimation: point and interval estimator
Confidence interval for Normally distributed mean:
population variance is known
We use basic algebra to find the following:
h i
Pr − zα/2 < Z < zα/2 = 1 − α
h x̄ − µ i
Pr − zα/2 < √ < zα/2 = 1 − α
σ/ n
...
h σ σ i
Pr x̄ − zα/2 √ < µ < x̄ + zα/2 √ = 1 − α
n n
For a 95%(α = 0.05) confidence level it follows that
h σ σ i
Pr x̄ − 1.96 √ < µ < x̄ + 1.96 √ = 0.95
n n
Estimation: point and interval estimator
Confidence interval for Normally distributed mean:
population variance is known
Consider a random sample of n observations from a normal distribution
with unknown mean µ and known variance σ 2 . If the sample mean is x̄,
then a 100(1 − α)% confidence interval for the population mean with
known variance is given by
σ
x̄ ± zα/2 √
n
with
σ
Lower confidence Limit (LCL) = x̄ − zα/2 √
n
σ
Upper confidence Limit (UCL) = x̄ + zα/2 √
n
Estimation: point and interval estimator
Example
Suppose that shopping times for customers at a local mall are normally
distributed with known population standard deviation of 20 minutes. A
random sample of 64 shoppers in the local grocery store had a mean time
of 75 minutes. Find the the upper and lower confidence limits of a 95%
confidence interval for the population mean, µ.
Solution:
σ
LCL = x̄ − zα/2 √ =
n
σ
UCL = x̄ + zα/2 √ =
n
Estimation: point and interval estimator
Confidence interval for Normally distributed mean:
population variance is unknown
Consider a random sample of n observations from a normal distribution
with unknown mean µ and unknown variance σ 2 . In practice population
variance might be unknown
In the case where the population standard deviation is unknown, Z
statistic (which follows standard normal distribution) cannot be used
directly. It is natural in such circumstances to consider the random variable
obtained by replacing the unknown σ by the sample standard deviation,
sP
(xi − x̄)2
s=
n−1
If the sample mean is x̄ then an alternative to Z statistic is given by
x̄ − µ
t = √ ∼ tn−1
s/ n
where n − 1 is the degrees of freedom.
Estimation: point and interval estimator
Confidence interval for Normally distributed mean:
population variance is unknown
Then a 100(1 − α)% confidence interval for the population mean µ with
unknown variance is given by
s
x̄ ± tα/2,n−1 √
n
with
s
Lower confidence Limit (LCL) = x̄ − tα/2,n−1 √
n
s
Upper confidence Limit (UCL) = x̄ + tα/2,n−1 √
n
where tα/2,n−1 is the value from t distribution with n − 1 degrees of
freedom.
Estimation: point and interval estimator
Example: Gasoline Consumption
Recently gasoline prices rose drastically. Suppose that a study was
conducted using truck drivers with equivalent years of experience to test
run 24 trucks of a particular model over the same highway. Estimate the
population mean fuel consumption for this truck model with 90%
confidence if the fuel consumption, in miles per gallon, for these 24 trucks
was as follows:
15.5 21.0 18.5 19.3 19.7 16.9 20.2 14.5 16.5 19.2 18.7 18.2 18.0 17.5 18.5
20.5 18.6 19.1 19.8 18.0 19.8 18.2 20.3 21.8
Estimation: point and interval estimator
Example: Gasoline Consumption
P xi
Solution: Sample mean x̄ = = 18.68,
qn P
(xi −x̄)2
Sample standard deviation s = n−1 = 1.69526 and
The values from t distribution: t23,0.05 = 1.714.
90% CI for true mean fuel consumption for all trucks of this model:
LCL = x̄ − tα/2,n−1 √sn = 18.1
UCL = x̄ + tα/2,n−1 √sn = 19.3
Interpretation: If independent random samples of 24 trucks are
repeatedly selected from the population and confidence intervals for each
of these samples are determined, then over a very large number of
repeated trials, 90% of these intervals will contain the value of the true
mean fuel consumption for this model truck
Estimation: point and interval estimator
Comments on CI
Wide confidence interval reflects uncertainty about the parameter
being estimated.
A larger sample size n will give a narrower interval.
