Physical Activity, Exercise Physiology, Movement
Analyzing Social Networks Using Spatial Statistical Methods
(School of Public Health (UMD) Epidemiology and Biostatistics Master's Student)
A major contributor to the obesity epidemic is the decline of physical activity in today’s society. Studies have shown physical activity to be influenced by friendship. The goal of this study is to determine whether adolescent girls’ physical activity levels are influenced by their social networks and to predict physical activity while accounting for the social network effect using Spatial Statistics methods. Using data from the Trial of Activity for Adolescent Girls II, social networks of 8th grade girls in six schools were recorded in adjacency matrices accounting for friendships where others nominated the ego (in-degree), the ego nominated others (out-degree), and ego and other nominated each other (symmetric). Mean minutes of moderate to vigorous physical activity (MVPA) was the measure of physical activity used. Moran's I analyses were conducted on the six schools combined and Permutation Test were conducted on schools individually to detect clustering of MVPA. Simultaneous Autoregressive (SAR) and Conditional Autoregressive (CAR) models were run to predict MVPA while controlling for BMI, Race, and School, and accounting for the clustering effect of the schools combined. Moran's I analysis revealed significant clustering for each type of friendship network (in-degree p-value:0.000001503, out-degree p-value:0.0005098, symmetric p-value:0.0006385). Individual school permutation tests showed significant clustering only in the out-degree matrix of school 6 (p-value: 0.046).The CAR and SAR models revealed that girls from schools 3,4, and 6 had MVPA levels significantly lower than school 1 (ranging between 4 and 5 units difference) while accounting for each friendship definition. Results revealed a community effect rather than a clustering of MVPA by friendship. Students in the same school had similar MVPA levels significantly different than other schools. The school itself may have a significant impact on physical activity levels, and future interventions should be school based to be most effective.