March 10, 2005 :: No. 147

What’s a scientist to do when some individuals under study for depression are so depressed they don’t participate?

Those missing study participants mean missing data, which could result in faulty analysis.

For Mortaza Jamshidian, associate professor of mathematics at Cal State Fullerton, the search for a way to resolve that dilemma is an ideal project for math students.

“Incomplete or missing data is a pervasive problem in almost all areas of research,” says Jamshidian. “When someone drops out of a research study randomly, if they move out of state or have an accident on the way to the test site, that is seen as data missing ‘randomly.’ But in the case of those dropping out of a study because their condition is so severe, or because they don’t have the condition at all, that is considered ‘non-random.’ That type of missing data, if ignored, can lead to biased conclusions in a study,” notes the researcher.

Thanks to a three-year, $168,541 National Science Foundation grant, Jamshidian —working with a graduate student and two undergraduates — is developing, advancing and applying statistical methods for analysis of social and behavioral sciences data that are missing “not at random.”

“We’re working in two parts. The first part is fairly theoretical — development of statistical methods, using a lot of mathematics and statistics. The next stage is implementing the methods and programming our methodology to test our theoretical developments,” Jamshidian says.

Another phase is using these statistical methods on real data. To do that, Jamshidian is working with Ke-Hai Yuan, University of Notre Dame associate professor of psychology, and applying the Cal State Fullerton methods on data from Notre Dame’s Adolescent Parenting Project, an ongoing longitudinal study of teen parenting. An additional $168,459 National Science Foundation grant was awarded to Notre Dame for this phase of the project, according to Jamshidian.

“One thing we do know is that there won’t be a single method that will resolve all missing data problems in social and behavioral sciences. Our hope is that we find a series of methodologies that, depending on the type of missing data, will be of help to researchers.”

Jamshidian joined Cal State Fullerton in 2002, after teaching at the University of Central Florida in Orlando and Isfahan University of Technology in Iran. He also has served as a statistician with BMDP Statistical Software in Los Angeles. The Anaheim Hills resident earned his doctorate in applied mathematics from UCLA.

« back to News Front