Page 185 - Contributed Paper Session (CPS) - Volume 4

CPS2169 Carmen D. Tekwe et al.
year, 24 teachers from eight elementary school were recruited and randomly
assigned to the use of either standbiased desks {Stand2learn LLC College
Station, TX, USA, stand-biased desk (models S2LK04) and stool (models
S2LS04)} or traditional desks (model 2200 FBBK Series by Scholar Craft
Products, Birmingham, AL), and chairs (9000 Classic Series, by Virco Inc.,
Torrance, CA, USA) for in-class activities (Benden, et al. 2014). A total number
of 374 students from second through fourth grades were assented and
included in the study at baseline. Each student's height and weight were
obtained at the start of each semester by trained research assistants to
calculate their BMI. The study participants were required to wear calibrated
BodyMedia SenseWear® armband devices (BodyMedia, Pittsburgh, PA)
during the school hours for a week for each semester from fall 2012 to spring
2014. The devices recorded subject-specific steps counts and caloric energy
expenditure per minute while worn. Of the 374 recruited students, 193
students completed the study, while the remaining either graduated from
elementary school or their parents retracted their consent from the study.
Students with large proportions of missing data were excluded from our
analyses. Thus, our final analytic sample size was 157. The study was approved
by the Texas A&M IRB. To analyse the data, we considered the linear
regression model (LRM), functional linear regression model (FLRM), and
conditional functional quantile regression model (CFQRM).

3. Result
The mean BMI at baseline (fall semester of year 1) was 17.14; kg/m2 ;
(SD=2.71), while the mean BMI at the end of the study (spring semester of
year 2) was 17.55; kg/m2 ; (SD=3.17). The study sample was composed of 77
girls and 80 boys and the average age of the enrolled students at baseline was
7.73; (SD = 0.74) years. About 75.8% of the students were whites, 7% Hispanics,
7% African Americans, and 10.2% Asians/native Americans. In Figure 2, we
provide density plots of the log(BMI) at 18 months post-baseline and the
residuals obtained after adjusting for baseline covariates and clustering. The
skewness and long tails of both distributions indicate a possible violation of
the normality assumption from linear regression models. Prior to fitting the
mixed effects model considered, we computed the log of BMI at both baseline
and 18 months post baseline. Through AIC comparisons, we determined that
the log transformation of BMI provided a better fit than the inverse and square
root transformations. Next, mixed effects models were used to obtain the
baseline and cluster randomized adjusted residuals for BMI. Overall, we did
not find a significant impact of age at baseline on the BMI values at 18 months
post baseline (p=0.209). However, there were statistically significant
differences between African Americans and whites (p= 0.05). The random

174 | I S I W S C 2 0 1 9