|
Abstract:
|
The use of data mining techniques to determine appropriate variables and interactions to investigate in a statistical or epidemiological model is demonstrated. The approach is compared with results from a model developed using more familiar techniques such as stepwise regression. Factors affecting the use of hearing protection in farmers were investigated. Age, years in farming, hearing threshold, use of noisy farm equipment, education, off-farm jobs, presence of deafness or tinnitus, training in occupational safety and health practices, and perceptions of usefulness of personal protective devices (PPD) were considered. Attitudes toward PPD, presence of tinnitus or deafness, and noisy off-farm jobs were most significant under both approaches. The data mining approach offers a graphical depiction of variables while also suggesting categorical breakdowns for continuous variables.
|
