Note on a Method to Estimate Missing Air Pollution Data

An iterative regression procedure is presented to estimate missing air pollution measurements when the data are measured at two or more sampling stations in the same vicinity. The procedure utilizes the measurements taken at other stations, on neighboring days, and of other pollutants.

The procedure is applied to a set of Philadelphia pollution data with from five to seventeen per cent of the observations missing. The method is tested by comparing the known observed pollutant values, with their estimates given by the procedure. Correlations between the observations and their estimates are uniformly high, ranging from 0.87 to 0.91. These correlations compare favorably with those estimates given by a simple linear interpolation. The magnitude of the correlations suggests that estimates given by this iterative regression procedure may be used where missing observations pccur without fear of undesirable effects on subsequent work. Therefore, this procedure may be a valuable tool in handling the problem of missing observations in air pollution data.