Statistical Analysis
We used the v2 test to determine whether
there was a difference in baseline demographic
and other characteristics between the study
and comparison groups (Table 2). Type of
residence was the only significant difference
between the 2 groups, with 32% of study
group children living in apartments compared
with 53% of comparison group children
(P = .049). Because type of home could in-
fluence the type of weatherization-plus-health
interventions conducted in a given dwelling,
we adjusted for these differences using propensity
score weighting, controlling for the
differences between the 2 groups; this resulted
in an unbiased estimation of the treatment
effect. To create the propensity score, we used
a logistic regression model to predict the
log-odds of being in the study group vs the
comparison group. The regression model
was based on child’s age (3---6 vs ‡ 7 years),
apartment versus house, winter (December
21---March 20) data collection period (yes vs
no), and year of construction (1940---1959,
1960---1979, or 1980---2009).