In an individual-level study, epidemiologists collect information -- disease outcome, exposure and covariates -- for each subject. Ecologic studies, in contrast, collect and analyze aggregate data. As an example of the latter, one might compare average meat consumption with colorectal cancer by country. (The terms "ecologic" and "ecological" are used interchangeably in this context, but neither refers to ecology: They refer to "groups" or grouped data.) Ecologic studies are attractive because they are often inexpensive and easily conducted, relying on routinely collected data. However, ecologic studies are subject to very large potential biases: relationships between group-level variables do not necessarily reflect relationships between individual-level variables and vice versa. Furthermore, many nominally individual-level studies employ one or more group-level variables, raising the potential of ecologic bias in such partially-ecologic studies. While research has traditionally focused on bias arising from estimating individual-level parameters from fully aggregated data, other study designs are receiving increased attention: partially-ecologic studies and studies with contextual (purely group-level) effects. The workshop will particularly focus on methods that combine individual and group level data. While most research on ecologic bias is theoretical, insight into the amount of ecologic bias actually occurring in real studies can be obtained from parallel analyses of the same data set on the individual and group levels. Both theoretical and applied papers are of interest. The workshop will bring together epidemiologists and biostatisticians/mathematicians to discuss these developments. For additional background on ecologic bias, please see: http://www.cireeh.org/pmwiki.php/Main/Ecologicstudies
Next: Call for Participation
Workshop Index
DIMACS Homepage