INTERVAL

ESTIMATION

Note: Most of the Slides were taken from

Elementary Statistics: A Handbook of Slide
Presentation prepared by Z.V.J. Albacea, C.E.
Reano, R.V. Collado, L.N. Comia and N.A.
Tandang in 2005 for the Institute of Statistics,
CAS, UP Los Banos
 TEACHING BASIC STATISTICS …

To introduce confidence interval

estimation….
◼ Subjective against statistical
confidence interval
◼ Confidence level
◼ Common features of all
confidence interval
estimates

Session 6.2
 TEACHING BASIC STATISTICS …

Subjective vs. Statistical

Confidence Interval
◼ A subjective confidence interval
is based on intuition.
◼ A statistical confidence interval
estimate is based on sample
values with an assumed
measure of confidence.

Session 6.3
 TEACHING BASIC STATISTICS …

Level of Confidence
◼ Denoted by 100 (1 −  ) %
◼ Typical levels: 90%, 95%, 99%
◼ A relative frequency interpretation
 In the long run, 100 (1 −  ) % of all the confidence
intervals that can be constructed will contain the
unknown parameter. (In other words, the chance of
selecting a sample that will yield an interval
containing the value of the parameter is (1-).)
 Wrong to say: that the computed interval will have (1-)
probability of containing the parameter.
◼ For a fixed n, the length of the interval increases when
we increase the value of (1-).

Session 6.4
 TEACHING BASIC STATISTICS …

Confidence Interval Estimation

◼ An ideal estimate is one that is
narrow and accurate.
◼ Interval estimation is based on
repeated sampling, but for a
given case, we usually have only
a single interval under
consideration.

Session 6.5
 TEACHING BASIC STATISTICS …

Confidence Interval Estimation

◼ The single confidence interval is either
correct or incorrect (all or nothing), but
the confidence level gives us an
indication of the proportion of correct
intervals that can be expected with
repeating the estimation procedure.
◼ Once an interval is constructed, we do
not find out if it is actually correct.

Session 6.6
 TEACHING BASIC STATISTICS …

Interpreting Confidence Intervals

◼ The 95% confidence

intervals are different for 100
samples
◼ For about 95 of them the
interval covers the
parameter, but about 5 do
not!

Session 6.7
 TEACHING BASIC STATISTICS …

Confidence Interval for m

(s is known)
◼ Assumptions
Population SD known
Population is normally distributed
◼ Confidence interval estimate
s s
X − Z / 2  m  X + Z / 2
n n

Session 6.8
 TEACHING BASIC STATISTICS …

Confidence Interval for m

(s is known)
X  Zs X = X  Z
s
n
sx_
_
X
m − 1 .645 s x m + 1 .645 s x
90% Samples

m − 1.96s x m + 1.96s x
95% Samples
m − 2.58s x m + 2.58s x
99% Samples

Session 6.9
 TEACHING BASIC STATISTICS …

INTERVAL ESTIMATION
◼ The general form of an interval estimate is
point estimate  maximum error
where maximum error = (critical value 
standard error of the estimate).
◼ Inestimating the mean, maximum error is
given by
s
error = Z
n

Session 6.10
 TEACHING BASIC STATISTICS …

Determining Sample Size (Cost)

Too Big:
Too small:
• Requires
too much • Won’t do
the job
resources

Session 6.11
 TEACHING BASIC STATISTICS …

Determining Sample Size for Estimating

the Population Mean

What sample size is needed to be 90% confident

of being correct within ± 5? A pilot study
suggested that the standard deviation s is 45.

Zs 2 2 1.645 ( 45
2 2
) = 219.2  220
n= 2
= 2
Error 5
Round Up
Session 6.12
 TEACHING BASIC STATISTICS …

Confidence Interval for m

(s is unknown, but large sample)
◼ Assumptions
PopulationSD unknown
Large sample size
◼ Confidence interval estimate
s s
X − Z / 2  m  X + Z / 2
n n

Session 6.13
 TEACHING BASIC STATISTICS …

EXAMPLE

The mean and standard deviation for the

quality point indices (QPI) of a random sample
of 36 students are 2.6 and 0.3, respectively.
Find the 99% CI for the mean.
Solution: 99% CI for the mean
0.3 0.3
2.6 − 2.575  m  2.6 + 2.575
36 36
or equivalently, 2.47 to 2.73

Session 6.14
 TEACHING BASIC STATISTICS …

Confidence Interval for m

(s is unknown)
◼ Assumptions
Population SD unknown
Population is normally distributed
◼ Confidence interval estimate
s s
X − t / 2,n −1  m  X + t / 2, n −1
n n

Session 6.15
 TEACHING BASIC STATISTICS …

Student’s t Distribution
Standard
Normal

Bell-Shaped t (df = 13)

Symmetric
‘Fatter’ t (df = 5)
Tails

Z
0 t
Session 6.16
 TEACHING BASIC STATISTICS …

Degrees of Freedom (df )

Number of observations that are
free to vary after sample mean has
been calculated degrees of freedom
= n -1
Example = 3 -1
Mean of 3 numbers is 2 =2
X 1 = 1 (or any number)
X 2 = 2 (or any number)
X 3 = 3 (cannot vary)

Session 6.17
 TEACHING BASIC STATISTICS …

Student’s t Table

Upper Tail Area Let: n = 3

df = n - 1 = 2
df .25 .10 .05  = .10
1 1.000 3.078 6.314 /2 =.05

2 0.817 1.886 2.920  / 2 = .05

3 0.765 1.638 2.353

t Values 0 2.920 t
Session 6.18
 TEACHING BASIC STATISTICS …

EXAMPLE

The contents of 8 similar bottles of

acetic acid are 110, 112, 111, 109,
107, 113, 110, and 109 milliliters.
Find a 95% confidence interval for
the mean of all such bottles,
assuming an approximate normal
distribution for the population of the
acetic acid contents.

Session 6.19
 TEACHING BASIC STATISTICS …

Example

◼ Here, we have:
x = 881/ 8 = 110.125
s = 3.554 = 1.89
◼ Thus the 95% confidence interval is:
1.89
110.125  2.365
8

or, equivalently, from 108.545 to 111.705.

Session 6.20
 TEACHING BASIC STATISTICS …

Confidence Interval for the

Difference of Two Means
◼ Parameter of interest is
m D = m1 – m2
◼ Assumptions
 Known SDs
 Normally distributed or large samples

◼ Confidence interval estimate

s 2
s 2

(x1
− x2 )  Z  / 2 1
+ 2

n1 n2
Session 6.21
 TEACHING BASIC STATISTICS …

Confidence Interval for the

Difference of Two Means
◼ Parameter of interest is
mD = m1 – m2
◼ Assumptions
Unknown but equal variances
Normally distributed or large
samples

Session 6.22
 TEACHING BASIC STATISTICS …

Confidence Interval for the

Difference of Two Means
◼ Confidence interval estimate
2 2
s pooled s pooled
(x
1
− x2 )  t (  / 2 , n + n − 2 ) +
1 2
n1 n2
where
(n1 − 1) s12 + (n2 − 1) s 22
s 2pooled =
n1 + n2 − 2

Session 6.23
 TEACHING BASIC STATISTICS …

Confidence Interval
Estimate for Proportion
◼ Assumptions
 Normal approximation can be used if
np  5 and n 1 − p  5 ( )
◼ Confidence interval estimate
pˆ (1 − pˆ ) pˆ (1 − pˆ )
pˆ − Z   p  pˆ + Z 
2
n 2
n

Session 6.24
 TEACHING BASIC STATISTICS …

EXAMPLE
A random sample of 400 voters showed 32 voted for
Drilon in the next presidential election. Set up a 95%
confidence interval estimate for p.
pˆ (1 − pˆ ) pˆ (1 − pˆ )
pˆ − Z  p  pˆ + Z
2
n 2
n
.08(1 − .08) .08(1 − .08)
.08 − 1.96  p  .08 + 1.96
400 400
.053  p  .107

Session 6.25
 TEACHING BASIC STATISTICS …

Determining Sample Size for Proportion

◼ Sample size formula

Z p(1 − p)
2
n= 2
Error
◼ When p unknown, either estimate
from pilot survey or use p=1/2

Session 6.26
 TEACHING BASIC STATISTICS …

Ethical Issues

◼ Confidence interval (reflects sampling

error) should always be reported along
with the point estimate
◼ The level of confidence should always
be reported
◼ The sample size should be reported
◼ An interpretation of the confidence
interval estimate should also be provided

Session 6.27