The paper focuses on the adjustment of the exposure indices developed in work on exposure-response studies on crops to deal with missing data patterns in ambient air quality data and present three statistical methods that are fundamentally related to the methods proposed by Davidson and Hemphill (1987) for dealing with entire days of missing data and incomplete days. These methods are applied to hourly O3 data at eight nonurban monitoring sites in the United States to compare the capabilities of each method to adjust for varying frequencies and temporal distributions of missing observations. Because most of the data on atmospheric pollutants and related variables available in the United States from continuous-monitoring networks has been collected at urban sites (Schere, 1988), the paper focuses on the use of statistical methods that do not require concomitant variables for dealing with missing observations.