Theory of Estimation

Statistical inference is the process that

involves using limited information obtained
from a sample to draw conclusions about
population from which the sample is taken.
It may take the form of developing estimates
about population characteristics or testing of
hypotheses about the characteristics of the
population.
Point Estimate example:
55 percent of the people in the country
favour the Indo-American nuclear deal. A
point estimate is most likely to provide an
inaccurate estimate.
Interval Estimate example:
50% - 60% of the people in the country favour
the Indo-American nuclear deal.
Interval estimate is expected to provide a
more dependable estimate. Wider the interval,
more the chances of being correct.
Estimation: The process where sample
information is used to estimate the numerical
value of some population measure such as
population mean, population variance.
Estimate: An educated guess about some
value of a population parameter.
Estimator : The rule or procedure used to
obtain that guess.
 An interval estimate gives the estimate in terms
of a range of values.

An interval estimate is obtained as under:

Point estimate ± (Interval co-efficient × Standard

error)
 Standard error is the standard deviation of the
sampling distribution in question,
 The interval co-efficient or confidence co-
efficient as it is called, depends on the level of
confidence and the shape of the sampling
distribution.
SAMPLING DISTRIBUTION OF MEAN AND
CONFIDENCE INTERVALS
 Interval estimate of µ may be constructed as:

Thus
 Upper Limit of Interval:
Lower Limit of Interval:
Where,

standard error of mean

value determined by probability
α is associated with level of confidence,
If level of confidence is 95%, then α = 0.05, and
α / 2 = 0.025.
,

95% confidence interval for mean

99% confidence interval for mean

Population Population Standard

Standard Deviation Deviation Not
Known Known

SE= SE = where
A marketing research firm has contracted with an
advertising agency to provide information on the
average time spent on watching TV per week by
families in a city.
The firm selected a random sample of 100 families
and found the mean time spent by them watching TV
to be equal to 32 hours. It is known that the standard
deviation of the TV watching time is 12 hours, which
has been constant over the past few years.
The firm wants to construct a 95 percent confidence
interval for mean time spent by families watching TV.
Help the firm.
 Here, n= 100 (sample of 100 families )
= 32, (the mean time spent watching TV)

For a 95 percent confidence interval,

α / 2 = 0.05 /2 = 0.025 and z 0.025 = 1.96
The 95 percent confidence interval for mean:

32 ± 1.96 × 1.2
29.648 - 34.352
So lower limit of the estimate is 29.648 hours and
the upper limit of the estimate is 34.352 hours.
 It is desirable that:
 The level of confidence should be high so that
the parameter value is contained within the
interval we obtain, and
 The width of the interval should be small. A
smaller interval is always more useful than a
larger one. This is because a wider interval
implies lesser precision of the estimate.
For a given sample size, higher the level of
confidence, wider the interval estimate,
While for a given level of confidence, larger the
sample size, the smaller the width of the class
interval.
Interval estimate of population mean:

So, there is a 1 – α probability that the value of

sample mean will provide a sample error of
or less
Given the extent of error acceptable (E), the level of
confidence (1 – α),

Rearranging the terms,

Sample size
Oil India Corporation has a bottle-filling
machine which can be adjusted to fill oil to any
given average level, but it fills oil with a
standard deviation of 0.010 litres.
The machine has recently been reset to a new
filling level and the manager wants an estimate
of the mean amount of fill to be within ± 0 . 001
litres with a 99 percent level of confidence.
How many bottles should the manager sample?
Let Erepresent the maximum amount of error.
Thus = 0.001. Also, σ = 0.010.
Corresponding to 99 percent level of
confidence, z = 2.58.
Now, with these inputs, we have

Thus, 666 bottles should be examined by the

manager.
 SMALL SAMPLES, n ≤30

When
•Population is normally distributed, and
•Population standard deviation is known

Use z -
statistic
 SMALL SAMPLES, n ≤30

•Population is normally distributed

•Population standard deviation is NOT known

Use t -statistic
A t distribution is similar to normal distribution in
many ways.
Like normal, a t distribution is symmetrical, bell-
shaped and extends from − ∞ to + ∞ .
Both the standardized distributions have mean
(median and mode) equal to zero.
The prime difference between a t and a z -distribution
lies in terms of the spread of the two.
Recall that the variance of a standard normal
distribution is equal to 1.0.
However, the variance of a t distribution is always
greater than 1.0. The variance of a t distribution is
given by ν/(ν – 2) (for ν > 2) or (n – 1)/( n – 3).
Thus, the exact variance of a t distribution
depends on the size of the sample and it
decreases as the value of n increases.
Since the variance decreases as the size of the
sample increases, it follows that the degree of
flatness decreases with increases in the sample
size n and it gradually tends to approach the z -
distribution.
Although t distribution is applicable in all cases
where the population standard deviation is not
known, we use z -distribution in its place when
samples are “large” in size.
 Based on
degrees of
Used for small samples freedom
The t variable is defined as:

where in s is the standard deviation of the

sample and is obtained as
Population mean can be estimated using t
distribution as:

OR

Where,
If we want to make estimate at 95 percent
level of confidence, the t -table should be
consulted for 5 percent.
Another element relevant in consulting the t -
table is the number of degrees of freedom, ν,
which is equal to n – 1 for a sample of size n.
A sample of 20 items is selected randomly
from a very large shipment. It is found to have
a mean weight of 310 gm and standard
deviation equal to 9 gm. Derive the 95 percent
confidence interval for the population mean
weight.

Note: Sample is small (≤ 30), and

population standard deviation is
not known. The weight is
expected to be normally
distributed.
So we would use t - statistic
 Here, n = 20, Degrees of freedom ν = n – 1 = 19,
sample mean = 310, and s = 20

95 percent confidence interval:

For ν = 19 and α = 0.05

or 305.788 to 314.212
Thus, there is 95 percent likelihood that the
population mean will be included in the interval
305.788 gm – 314.212 gm.
 Shape of ParentPopulation
Sample size
Not‐normally Distributed
Normally Distributed
(Skewed)

(σ known) (σ known)
Large

(σ unknown) (σ unknown)

(σ known)
Small No intervalpossible

(σ unknown)
.

The proportion of successes in the population, π,

is given by the ratio X/N,
Where X is the total number of successes in the
population and N refers the size of the
population.
p is a point estimate for the proportion of
successes in the population, π .
S.E is
The confidence interval can be obtained as:
Of the 200 people selected randomly from a
large city, 120 were found to be in favour of a
more liberal divorce law.
Using this information, obtain a 95 percent
confidence interval of the proportion of people
in the city who favour more liberal divorce law.
n = 200 and p = 120/200 = 0.6.

Further,

95 percent confidence interval :0.6 ± 1.96 × 0.035

0.533 and 0.667.

It means that the chance is 0.95 that the interval

0.533 – 0.667 contains the proportion of people in the
city who prefer a more liberal divorce law.
Sample size required to achieve the given level of
confidence and the amount of tolerable error :

Estimated value of π can be used from pilot

sample.
Use π = 0.5 if no information is available.
Taking this value of π yields the highest value of
the product π (1 – π ), and, therefore, the largest
value of the standard error σ p .
There are two populations with means µ 1 and
µ 2 , and standard deviations as σ 1 and σ 2 .
From the population 1, select a random
sample of size n1
From the second population, select a random
sample of size n2.
Te estimate:
We use,
(

Sampling distribution of difference between

sample means has expected value
E( = , and
Standard error of difference between means:

Confidence interval of the difference between

means is given as
If population standard deviations are not
known,
We calculate the variance, s2, for each sample
The confidence interval is computed as:
When we consider two population means
using small samples with sizes of n 1 and n 2
respectively (so that n1 < 30 and n2 < 30)
The variances of the two populations are not
known, then
We first estimate variance of pooled sample:

OR
The confidence interval for difference between
population means as follows:

where
Sample mean from Population 1 & n1 its size
Sample mean from Population 2 & n2 its size
is the point estimate of the difference
between population means
t value for n 1 + n 2 – 2 degrees of freedom
is pooled variance
If populations having proportion of successes as π 1
and π 2 respectively, and samples, sized n1 and n2 ,
(n1 > 30 and n2 > 30) are selected,
The sampling distribution of difference between
sample proportions (p1 – p2)
E(Δp) or E(p1 – p 2) = Δπ
S.E.=

Using estimates from samples:

Confidence interval estimate of the difference
between population proportions :

OR