NAME:  Simpson's Paradox: An Example From a Longitudinal Study in South
Africa
TYPE:  Census
SIZE:  1590 observations, 3 variables

DESCRIPTIVE ABSTRACT: 
The dataset contains information for the construction of a three-way
table that illustrates Simpson's Paradox.  Categorical variables
observed for babies born in a metropolitan area of South Africa include
race, whether or not the mother had medical aid (which is similar to
health insurance), and whether or not the child was included in a
follow-up study.  The data are provided in two formats.  The file
birthtotena.dat contains the category labels and cell frequencies for
the three-way table.  The file birthtotenb.dat lists each case on a
separate line with three variables that indicate whether or not the
mother had medical aid, whether or not the mother was traced for the
five-year interview, and race.

SOURCES: 
The Birth to Ten data were obtained from the Chronic Diseases of
Lifestyle Programme at the Medical Research Council in Cape Town, South
Africa.  The author assisted with the analyses on the five-year
follow-up data.

VARIABLE DESCRIPTIONS: 
Key to Variables in birthtotena.dat
Columns
 1 -  5  Aid/NoAid
 7 - 15  Traced/NotTraced
17 - 21  White/Black
23 - 25  Cell Count

Values are aligned and delimited by blanks. There are no missing values.

Key to Variables in birthtotenb.dat
Columns
   1     Medical Aid? (0 = No, 1 = Yes)
   3     Traced? (0 = No, 1 = Five-Year Group)
   5     Race (1 = White, 2 = Black)

Values are aligned and delimited by blanks. There are no missing values.

STORY BEHIND THE DATA:  
The Birth to Ten study (BTT) commenced in the greater
Johannesburg/Soweto metropolitan area of South Africa during 1990.  A
birth cohort was formed from all singleton births during a seven-week
period between April and June 1990 to women with permanent addresses
within a defined area.  Identification of children born during this
seven-week period and living in the defined area took place throughout
the first year of the study, by the end of which 4029 births had been
enrolled.  The BTT study collected prenatal, birth, and early
development information on these children.  The aim of the study was to
identify factors related to the emergence of cardiovascular disease
risk factors in children living in an urban environment in South
Africa.  In 1995, when the children were five years old, the children
and caregivers were invited to attend interviews.  Detailed
questionnaires were completed that included questions about living
conditions within the child's home, the child's exposure to tobacco
smoke, and additional health-related issues.  The five-year sample
consisted of 964 children.  Unfortunately, there was a great deal of
missing data in the baseline group, especially on the variables
reported here.

If the five-year sample is to be used to draw conclusions about the
entire birth cohort, the five-year group should have characteristics
similar to those who were not traced from the initial group.  Thus, the
five-year group was compared to those who did not participate in the
five-year interview on a number of factors.  One of the factors was a
variable that determined whether the mother had medical aid (which is
similar to health insurance) at the time of the birth of the child.

Additional information about these data can be found in the "Datasets
and Stories" article "Simpson's Paradox: An Example From a Longitudinal
Study in South Africa" in the _Journal of Statistics Education_
(Morrell 1999).

PEDAGOGICAL NOTES: 
This dataset provides an interesting real-world example of Simpson's
Paradox.  The association between the medical aid variable and
five-year follow-up variable changes when white and black participants
are considered separately.

REFERENCES:
Levitt, N. S., Steyn, K., De Wet, T, Morrell, C. H., Edwards, R.,
Ellison, G. T. H., and Cameron, N. (1999) "An Inverse Relationship
Between Blood Pressure and Birth Weight Among 5 Year Old Children from
Soweto, South Africa," _Journal of Epidemiology and Community Health_,
53, 264-268.

Yach, D., Cameron, N., Padayachee, N., Wagstaff,  L., Richter, L., and
Fonn, S. (1991), "Birth to Ten: Child Health in South Africa in the
1990s. Rationale and Methods of a Birth Cohort Study," _Paediatric and
Perinatal Epidemiology_, 5, 211-233.

SUBMITTED BY: 
Christopher H. Morrell
Mathematical Sciences Department
Loyola College
4501 North Charles Street
Baltimore, MD 21210-2699
chm@loyola.edu