Collecting Data

1

• Collecting Data

A S 11.4.1
2
AFRICAS SILENT CRISIS The number of AIDS
orphans in sub-Saharan Africa alone was estimated
to be around 12 million in 2005 and is predicted
to increase to over 18 million in
2010. (UNICEF,2004UNAIDS,2006). It is estimated
that, in these countries, approximately 9 of
children under the age of 15 have lost at least
one parent to AIDS and one in six households with
children is caring for at least one orphan.
(Monasch and Boerma, 2004 UNAIDS, 2006).
3
Power from Data
To start the process of any statistical research
or investigation, a suitable set of data needs to
be collected to answer specific questions, as in
the article, before any informed decisions can be
made.
Data collection has to be carefully planned, else
the results may be incorrect or unreliable.
4
Sampling Bias
When doing a sample survey only part of the total
population is approached for data. Using the
whole population is impractical because of time
and cost.
Sampling errors (bias) can arise, by over
representation or under representation, when
using only a portion of the population rather
than the entire population.
5
A sample must be representative of the entire
population to which the data applies.
To Do

• What might be wrong in using the following
  samples to describe the South African people
• as a whole?
• Teenagers.
• People whose names begin with P.
• People using OR Tambo Airport.
• People in restaurants.
• Every tenth person in the telephone directory.

6
Avoiding Sample Bias

• Select the members randomly. Every member of the
  population under investigation must have an equal
  chance of being chosen.

e.g. Don't just ask people at the taxi rank when
trying to find out about how people get to
work.

• Use stratified sampling. Divide populations into
  sections and then randomly select the same
  proportion of members from each section.

e.g. Include the same proportions of
males/females, age groups, ethnic groups
etc. as in the target population.
7
A Principal wants to select a stratified sample
of 50 learners from Grade 10, 11 and 12. The
table show the numbers in each grade. How many
from each grade must be selected.
Learner total 320300180 800 Grade 10
320/800 x 50 20 pupils Grade 11 300/800 x 50
18,75 19 pupils Grade 12 180/800 x 50
11,25 11 pupils
8
Other Bias
A survey often asks participants to respond. If
they do not answer all the questions honestly and
accurately then the responses are biased.
Non-respondents (who will always be the
overwhelming majority) can be assumed to have no
decided view. In ignoring the undecided view, the
result is biased.
Questions must be clear, unambiguous and simple
to answer, else it could mean different things to
different people .
9
Whichever style (or styles) of questions you use,
it is important that they are unbiased.
To Do

• Criticize the following questions for bias.
• How would you improve the question to reduce
  bias?
• Do you think that the colour red is better
• than blue?
• Shouldnt the Government spend more on
• education than on defence?
• Don't you agree that this supermarket offers
• a great service?

10
Criticize this questionnaire
Are you concerned about the environment? Are you
concerned about the level of pollution caused by
cars? Is there too much congestion in the town
centre? Do you think the exhaust fumes are a
health risk to children? Should cars be banned
from the town centre?
Questionnaire very biased, leads one to answer
yes to last question. No questions about the
advantages of cars, methods to reduce pollution
or congestion etc.
11
T A S K
Your school is considering abolishing
part of its uniform, but wants to find the views
of all learners. Design a suitable questionnaire
of no more than 6 questions for this purpose. You
must collect information about the age, gender
and grade of the learner and two different
related issues e.g. what of time do you wear
your uniform?