You are on page 1of 9

A research is being done among cellphone users and let us assume, the mar

It is a customer satisfaction study, so satisfaction on 30-35 parameters are be

Findings of the study are required to be reported for both the brands by
- At "All Users" level
- For pre-paid user and post-paid user of Airtel and Vodafone separately

Our client has given two conditions for sample size determination:
1. Post-paid users of each are more important - so, there can be 4% error per
2. For pre-paid users of each, 5% error is permissible at 95% conf level

Qa) What will be the sample size of this research?

Qb) If the population of users is such that Airtel has 60% share and Vodafone
and, both have a 30-70 break up by Post-Paid and Pre-Paid
What weights will you use to report findings at "All Users" level?

Solution: Part a)
So, there are 4 sub-groups in which findings are to be reported
Post-Paid Pre-Paid
Airtel 1 2
Vodafone 3 4

So, let us calculate the sample sizes for the Cells 1 and 3

e= Zvalue X Sqrt(pq/n)
Here,
e= 0.04
z= 1.96 (at 95% conf)
p and q= 0.5 (where the error is highest, to be taken as the

0.04 X 0.04 = 1.96 X 1.96 X 0.5 X 0.5


n
n= 600
So, the min sample size required for Cell 1 and Cell 3 are 600 each

Similarly for Cell 2 and Cell 4

e= Zvalue X Sqrt(pq/n)
Here,
e= 0.05
z= 1.96 (at 95% conf)
p and q= 0.5 (where the error is highest, to be taken as the

0.05 X 0.05 = 1.96 X 1.96 X 0.5 X 0.5


n
n= 385

So, the min sample size required for Cell 2 and Cell 4 are 385 each

Thus, to fulfil the conditions of the client, the basic requirement of sample size

Post-Paid Pre-Paid Total


Airtel 600 385 985
Vodafone 600 385 985
Total 1200 770 1970

Thus, a total sample size of 1970 will be required

Solution b)
Now, the population is such that Airtel has 60% share and Vodafone 40%
And, both have a 30-70 break up by Post-Paid and Pre-Paid

In that case the population proportions become as follows:

POP %s Post-Paid Pre-Paid Total


Airtel 18% 42% 60%
Vodafone 12% 28% 40%
Total 30% 70% 100%
In addition, let us look at the sample sizes and proportions:
SAMPLE SIZE Post-Paid Pre-Paid Total
Airtel 600 385 985
Vodaphone 600 385 985
Total 1200 770 1970

So, the Sample proportions are as follows:


SAMPLE %s Post-Paid Pre-Paid Total
Airtel 30.5% 19.5% 50%
Vodaphone 30.5% 19.5% 50%
Total 61% 39% 100%

So, for reporting at 'All User' level, a weighting factor of Pop% / Sample% ha
The weighting factors would be as follows:
POP% / SAMPLE% Post-Paid Pre-Paid
Airtel 0.59 2.15
Vodaphone 0.39 1.43

WEIGHTING, A RATIONALE

Suppose you have done a study for FMCG among general mass with a sample size of 1
70% males and 30% females, maybe, because Males were more important as a group f

But the population is 50% males and 50% females.

So, while reporting findings at the 'All Market' level, if you use your sample proportions, y
by what the males have responded!!
Thus, you need to use weighting factors while analysing.

i.e. Females should get a multiplier (weighting factor) of 50/30 = 1.67 and males 50/70
s assume, the market has 2 service providers - Airtel and Vodafone
parameters are being studied for reporting

oth the brands by Post-Paid users and Pre-Paid users separately

ne separately

termination:
an be 4% error permissible at 95% confidence level
5% conf level

hare and Vodafone 40%

ers" level?

reported

to be taken as there are several, 30-35 parameters being studied)


re 600 each

to be taken as there are several, 30-35 parameters being studied)

re 385 each

ment of sample size is as follows:

and Vodafone 40%.


op% / Sample% has to be applied to the cells in the sample

ith a sample size of 1000, in which thr are 700 males and 300 females, i.e.
mportant as a group for this category and required more sub-groups to be represented by

sample proportions, you would over-represent males. Your findings will be skewed
67 and males 50/70
e represented by

be skewed

You might also like