Understanding Human Behavior
Understanding Human Behavior
It’s a challenge to tease out the relationships between behavioral influences. Take smoking, for example. Smokers often say they smoke to steady their mood. Until a few years ago, however, researchers could not verify whether this stabilization actually occurred because their statistical methods could not model variances. Methods existed for measuring averages in mood — whether someone felt generally happy or generally sad — but not its vacillation.
That is, a method to trace mood volatility did not exist until a University of Illinois at Chicago researcher developed a statistical method to address this question in a longitudinal study of smoking patterns among adolescents. That researcher, Donald Hedeker, professor of biostatistics in the UIC School of Public Health, recently developed software that will allow data analysts and social scientists to use his methodology.
Hedeker’s new program, named MIXREGLS and written in Fortran, is the location scale model of a methodological software package he created in the 1990s.
The free program is available at http://bit.ly/mixregls. The program file is bundled with a manual that illustrates two examples, the data sets for those examples and their syntax files. A paper describing the program and the examples, now in press at the Journal of Statistical Software, is also included.
With Robin Mermelstein, IHRP director and principal investigator of the longitudinal study that required the method, and another colleague, he first described the model supported by this new software in 2008 in the journal Biometrics.
The longitudinal study included ecological momentary assessment of 461 teens who used handheld computers to complete 30 or more surveys over a week, four times over two years. Each survey provided 30 to 40 observations about each teen’s mood, activity, behaviors and social interaction — a rich source of data with which to examine the relationship between mood and smoking.
“We asked, is it the case that as a person increases their level of smoking, their mood variance diminishes, gets more level?” he explained. “That’s exactly what we find in analyzing our data. This kind of methodology allows us to do that, whereas more standard statistical methods don’t have that capability.”
They reported their initial findings using this new method in the February 2009 issue of Addiction, concluding: “Following smoking, adolescents experienced higher positive affect and lower negative affect than they did at random, non-smoking times. Our analyses also indicated an increased consistency of subjective mood responses as smoking experience increased and a diminishing of mood change.”
In the journal Statistics in Medicine, published online in March 2012, they describedfurther how covariates influence mood variances and extended the statistical model “by adding a subject-level random effect to the within-subject variance specification.”
Hedeker, who is a member of the IHRP Methodology Research Core, has spent much of his career making statistical tools more accessible to data analysts and researchers.
He said, “In statistics, if you develop a methodology but don’t provide a way for people to use it, well…” He shrugged. “No one uses it.”
Hedeker developed the software program at the UIC Institute for Health Research and Policy with support by the National Cancer Institute of the National Institutes of Health (grant no. R21CA140696).
He directs the Center for Biostatistical Development in the Division of Epidemiology and Statistics in the UIC School of Public Health.