METHODOLOGY by monkey6

VIEWS: 147 PAGES: 7

More Info
									RESEARCH METHODOLOGY

DESCRIPTIVE / QUANTITATIVE RESEARCH
This chapter covers the planning and execution of the quantitative phase of this study. It covers method
of administration, sample design, questionnaire design and reference to the questionnaire itself.


METHOD OF ADMINISTRATION
A random stratified cluster sample was drawn from the sample frame, which consisted of all the
townships in the Cape Metropolitan Area. Nine female, Xhosa-speaking interviewers conducted the
personal interviews. The interviewers had little or no experience in the field and so a period of one week
was dedicated to training before the start of the fieldwork. The pilot study also helped iron out any
problems the fieldworkers had in understanding and completing the questionnaires, and in finding the
marked plots on the street maps.




SAMPLE DESIGN

SAMPLE FRAME
The sample frame included all black women living in Cape Town area townships who were between the
ages of 15 and 64 years. The entire sample was split evenly between the three housing contexts, namely:
formal, serviced areas; informal, serviced areas; and informal unserviced areas. Quotas were also
implemented to ensure an even age group split, as well as to ensure that a substantial amount of smokers
and snuff users were interviewed. Interviewers were instructed to find a respondent who matched the
description on their cover sheet, starting from a site chosen at random.


It is very difficult to obtain completely up to date information and maps of the township areas, but
extensive efforts were made to get the most recent site plans available. The most recent schedules of
occupied sites and squatter density for Khayelitsha were obtained from the Tygerberg office in
Khayelitsha, which were published in November 1994, and the site plans/maps for the other townships
were obtained from the Geographical Information Services (GIS ) and the Cape Town Town Planning
departments.


Data pertaining to age, gender, population and services per area, was obtained by special request from the
1996 Census, completed by Statistics South Africa (SSA). We based our sampling on this population
data, and used other available sources of information for areas which were not covered by the census
(possibly because the settlements were new or more developed).
Below is the list of the areas that were chosen for the study using a random stratified cluster sampling
technique. They have been divided as best as possible into the three housing context, although there are
always exceptions within each area. It must be noted that this is the final list of areas in which we
interviewed. As the study progressed it was discovered at times that an area to be visited was no longer of
the housing context under which we had classified it. For example, certain areas classified as unserviced
areas in the 1996 census, were now serviced three years later. In cases such as these, it was necessary to
replace the area with another unserviced area on our list. This was always done in a completely randomly.


The tables from which we derived this data are available in Appendix H.


Suburbs To Be Included In The Final Study:


Formal, Serviced Areas:


1. Guguletu 1
2. Makhaya, Mandela Park - Khayelitsha
3. Guguletu 2
4. Section I & J - Khayelitsha
5. Guguletu 3
6. New Crossroads
7. Nyanga
8. Old Crossroads


Informal, Serviced Areas:
1. Harare - Khayelitsha
2. Macassar - Khayelitsha
3. Victoria Mxenge, Site B - Khayelitsha (50%)
4. Site C, C & D - Khayelitsha
5. Trevor Vilakazi - Khayelitsha
6. Town 2 / Griffiths Mxenge - Khayelitsha
7. Browns Farm 1
8. Browns Farm 2


                                                      2
Informal, Unserviced Areas:
1. New Rest - Guguletu
2. Victoria Mxenge, Site B - Khayelitsha
3. Trevor Vilakazi, Site B - Khayelitsha
4. Langa
5. Mkonto Square - Nyanga
6. Ktc- Nyanga
7. Greenpoint - Khayelitsha
8. Chris Hani - Khayelitsha (Buffer Around Site C)


The total female black population included in the sampling frame was 216454, out of which we
interviewed 1314.



SAMPLE SIZE
The sample size for the pilot test was 162 interviews. This was divided into 54 initial interviews to test
the questionnaire, another 54 to test interviewer reliability, and a further 54 to check the final revised
questionnaire.


The sample size for the main study was initially 1320 although the cleaned dataset consisted of only 1314
respondents. As has already been mentioned, the final sample was divided equally into three housing
contexts: formal serviced (440); informal, serviced (440); and informal, unserviced (40). Within each
housing context, the sample size was divided into two age groups: 15 - 29 years; and 30 - 64 years. In the
1998 pilot study the age group split was 15-34 and 35-65. This was however, revised in 1999 due to the
new split being a far more accurate split in terms of population per age group according to the 1996
census



SAMPLING PROCEDURE
A random method of selection was used. Stratified sampling within cluster sampling was performed in
deciding which suburbs to draw the respondents from. We interviewed 55 respondents from each of these
suburbs.


The total sample is divided as follows into the three quotas:
                                                                 312 smokers quota
                                                                 336 snuff users quota

                                                       3
                                                               672 random selection


Within each housing context (440 per housing context), the sample is divided as follows:


                                  104 smokers by quota (13 per suburb)
                                  112 snuff users by quota (14 per suburb)
                                  224 randomly (28 per suburb)


Nine fieldworkers were employed and trained. They were divided into three teams, which were changed
in their make-up three times during the fieldwork to try and minimise collaboration. There was a team of
two women the smoking quota, a team of three women for the snuff quota, and a team of four women for
the random quota. Collectively, the fieldworkers completed 55 interviews per day in the chosen suburb
per day.


The method which we adopted to draw our sample was as follows:
First we arranged the suburbs under the three housing context labels. Then, the suburbs of each housing
context were arranged into lists and a cumulative column was made of the populations in each suburb. A
list of random numbers, available in Appendix B, was used to choose eight (8) suburbs from each
housing context. In order to make up the sample of four hundred and forty (440) in each housing context,
fifty-five (55) respondents from each randomly chosen suburb were to be interviewed.


In order to randomly choose the plots to be marked as starting points on the maps, the number of erfs /
plots in each area were divided by the number of interviews per quota.(x : smokers = 13, snuff users = 14,
random = 28). (i.e. if there were n plots then every n/13th plot would be chosen as a starting point for a
smoking quota interview). Then it was just a matter of counting through the plots starting from one end
and systematically working towards the other, and marking off every n/xth plot. In order to implement the
quotas correctly and randomise the sampling.


Cover sheets were also made for each questionnaire, each of which contained a different set of random
numbers used for choosing which building to approach on the plot and which respondent to interview if
more than one was eligible. The exact description (age and tobacco usage status) of the respondent to be
found was printed on the cover sheet. Examples of these can be found in Appendix C. As plots were
randomly marked on the maps, the addresses or plot numbers were written straight onto the cover sheets,
ensuring that every second respondent chosen would be from a certain age group. Thus, both age and
housing contexts would be well randomised.


                                                      4
Photocopies were made of the maps with the marked plots on them and distributed to the fieldworkers.
Cover sheets were attached to clean copies of full questionnaires and individually placed in plastic
sleeves for cleanliness and safety reasons. Fieldworkers were given their questionnaires and incentives
for respondents on a daily basis, and this distribution was administered by the driver.


INCENTIVES
We were able to thank respondents for their participation, by giving them a perspex key-ring we had
specially manufactured for the study. The key-rings had the following message on the inside: (These
could be opened and replaced with a photograph if the respondent desired.)




QUESTIONNAIRE IMPROVEMENTS
The original questionnaire used in the pilot study in 1998 was 29 pages long. This proved far too long to
maintain interest from the respondents and so efforts were made to cut the 1998 questionnaire drastically
and to insert more new sets of questions too. The 1999 questionnaire is between 18 and 21 pages in total
including all the optional questionnaires (smoking, snuff, non-smoking), depending on the quota.




FIELDWORK
Initially nine fieldworkers were hired and trained for the project but only eight remained with us until the
end of the project. One of the fieldworkers' questionnaires proved unreliable and she was dismissed half
way through the fieldwork. Her questionnaires were re-done by three of the most reliable fieldworkers
after her dismissal.




                                                     5
A driver was hired to collect the fieldworkers from their homes every morning, take them to the
designated area for the day, and collect them at the end of each day and return them to their homes. The
driver was also responsible for collecting the clean questionnaires each evening from the Business School
and delivering the completed questionnaires at the end of each day to the fieldwork supervisor who
checked them and coded them.


Once in the designated suburb, the interviewers had to find the chosen plots on foot. The distances in
some cases were quite far and it was also difficult in certain suburbs for the interviewers to find the right
plot / site. This is because some of the maps obtained only erf / site numbers and not actual addresses,
which meant that the fieldworkers had to start at one end of the street and count the houses until they got
to the site that they calculated to be the right one.


Call-backs were conducted at the end of each day and interviewers were instructed to call-back twice
before finding another eligible respondent.


The interviewers were all trained for one week in total and were briefed again after the pilot test to point
out common errors and clear up any problems encountered in the field. The fieldwork was originally
expected to take six weeks to complete including the training and pilot test, but due to the unexpected
delays it took two months from start to finish. There are copies of the time-tables set at the back of the
Fieldworker's Manual which can be found in Appendix E. The manual gives all details of the fieldwork
and methodology followed, as well as all the suburbs from which respondents were drawn.


The fieldwork supervisor that was hired is the same woman who worked on our research project in 1998.
She is Patricia ----, a retired nurse currently living in Langa. She proved once again to be an invaluable
asset, especially in the training and translating arenas.



EDITING, CHECKING AND CODING
All the questionnaires were edited, checked and coded by the fieldwork supervisor. Not only did she
check that all relevant questions had been answered but she also made an effort to detect incorrect and/or
inconsistent answers and to complete any incomplete answers where the question tied in with other
completed questions. She was also responsible for ensuring that the fieldworkers had translated the open-
ended questions correctly from Xhosa into English.




                                                        6
DATA CAPTURE AND STATISTICAL ANALYSIS
The questionnaires were then checked again by the research project manager and the data was captured
in-house by an assistant hired for three months during the fieldwork stages of the project. The data was
cleaned by the project supervisor.



STATISTICAL ANALYSES
SPSS is the package that was mainly used for descriptive analyses and the Conversion Model was applied
to our research by Research Surveys Conversion Model Department. Our analyses has hung heavily on
the backbone of these results as they are seen to be very useful and accurate.




                                                     7

								
To top