It is
possible the associations detected in this analysis
may be distortions of the real underlying relationships
if the self-reported data are misrepresentations
of actual behaviors. However, variations of
the brushing variable, such as dichotomizing the
rate into once or twice a day or more, were tested in
the multivariable models (analyses not shown) and
generally yielded the same findings as models with
the continuous variable. The same pattern of
significant associations was also found when the
most extreme values (mothers who reported that
the child brushed 30–40 times per week) were
dropped from the continuous brushing frequency
variable.