"Software Glitch Threw Off Mortality Estimates," by Jocelyn Kaiser. Science, 14 June 2002, pages 1946-7;
"Statistical error leaves pollution data up in the air," by Jonathan Knight. Nature, 13 June 2002, page 677.
Both articles concern an error in a statistical software application that caused an overestimate of deaths that result from airborne soot. The fault is in the General Additive Model of S-Plus, a product of Insightful. Researchers used the model's iterative process to remove effects of other factors to estimate the effect airborne soot had on the increase per day in human mortality. But the program's default setting caused the iterations to terminate prematurely because the rise in mortality per day was so small. Industry groups are calling for a reexamination of Environmental Protection Agency air pollution rules. EPA officials and scientists have reexamined and are reexamining studies, and insist that almost all conclusions are not affected.
--- Mike Breen