diet {lcca} | R Documentation |
This dataset represents a simulated observational study to assess the effect of dieting on emotional distress among adolescent girls. This sample was drawn from an artificial population of one million girls described by Schafer and Kang (2008).
Variables in this population resemble actual variables from the first two waves of the National Longitudinal Study of Adolescent Health (Add Health) (Udry, 2003). However, no actual data from any Add Health participant appears in the population or in the sample; all data were randomly generated from probability distributions as described by Schafer and Kang (2008).
diet
a data frame with 6,000 rows and 17 variables:
DISTRESS.1
Emotional distress score at Wave I
BLACK
1=Black, 0=otherwise
NBHISP
1=non-Black Hispanic, 0=otherwise
GRADE
Grade in school at Wave I (7, ..., 11)
SLFHLTH
Self-rating of overall health (1=excellent, 2=very good, 3=good, 4=fair, 5=poor)
SLFWGHT
Self-rating of weight (1=very underweight, 2=slightly under, 3=about right, 4=slightly over, 5=very over)
WORKHARD
“When you get what you want, it's usually because you worked hard for it” (1=strongly agree, ..., 5=strongly disagree)
GOODQUAL
“You have lots of good qualities” (1=strongly agree, ..., 5=strongly disagree)
PHYSFIT
“You are physically fit” (1=strongly agree, ..., 5=strongly disagree)
PROUD
“You have a lot to be proud of” (1=strongly agree, ..., 5=strongly disagree)
LIKESLF
“You like yourself just the way you are” (1=strongly agree, ..., 5=strongly disagree)
ACCEPTED
“You feel socially accepted” (1=strongly agree, ..., 5=strongly disagree)
FEELLOVD
“You feel loved and wanted” (1=strongly agree, ..., 5=strongly disagree)
DISTRESS.2
emotional distress score at Wave II
U.1
first binary indicator related to dieting
U.2
second binary indicator related to dieting
U.3
third binary indicator related to dieting
Dieters and nondieters comprise about 20% and 80% of the population,
respectively. The treatment variable, a binary indicator of dieting
at Wave I, was removed
from the dataset and replaced by three conditionally independent binary
indicators (U.1
, U.2
, U.3
)
with endorsement probabilities of 0.90, 0.85 and 0.80 for
dieters and 0.10, 0.15 and 0.20 for nondieters. This
structure can be seen by fitting a latent class model with two classes
to U.1
, U.2
, and U.3
.
The response variable, DISTRESS.2
, is a simulated measure of
emotional distress at Wave II.
The remaining variables represent confounders recorded at Wave I which
influence girls' emotional distress and their propensities to
diet. Adjusting for these confounders is essential for estimating
causal effects of dieting on emotional distress. In particular, it is
essential to control for DISTRESS.1
, because this measure is
strongly related to dieting and to DISTRESS.2
.
The true causal effects in this population are very small. Dieting versus nondieting increases emotional distress by an average of 0.00253 for the entire population, by -0.02233 for dieters, and by 0.00859 for nondieters.
Schafer, J.L. and Kang, J.D.Y. (2008) Average causal effects from observational studies: a practical guide and simulated example. Psychological Methods, 13, 279-313.
Udry, J.R. (2003) The National Longitudinal Study of Adolescent Health (Add Health), Waves I and II, 1994-1996; Wave III, 2001-2002 (machine-readable data file and documentation), Chapel Hill, NC: Carolina Population Center, University of North Carolina at Chapel Hill.
For example analyses of this dataset using functions in
the LCCA package, see the manual LCCA Package for R, Version 1
in the subdirectory doc
.