Let's first consider measurement error in the exposure, PM2.5. Recall that measurement of PM2.5 exposure changed over time, due to differences in air quality monitoring data before versus after 1999. Consider the following hypothetical data, cross-tabulating PM2.5 levels measured in 2000 using higher quality monitoring data available post-1999, compared to PM2.5 levels measured in 2000 using lower quality monitoring data (estimated using more limited data). In this simplified scenario, we categorize PM2.5 as low (=20 ug/m3). Higher quality data High PM2.5 Low PM2.5 High PM2.5 70,000 60,000 quality Lower data Low PM2.5 5,000 280,267