AUTOCORRELATION OF INTERDAY EXPOSURES AT AN AUTOMOBILE ASSEMBLY PLANT

This article presents an analysis of interday autocorrelation in the e xposure time series of workers in an automobile assembly plant. In all , serial 8-hour time-weighted average exposures for six workers, two e ach at th ree separate processes, were monitored for periods ranging f rom 20 to 110 consecutive days. Autocorrelation was assessed in each c ase. In general the results support the conclusions of previous invest igations, in that relatively little autocorrelation was observed. Howe ver, uncertainties regarding the stationarity of the time series over the period of interest highlight potential difficulties in autocorrela tion analysis. Particularly, nonrandom events in a plant, such as vent ilation and process changes and extended down limes, should be conside red in any autocorrelation analysis inasmuch as these factors may affe ct the distributional parameters as well as the stationarity of the ti me series. In the workplace under study these nonrandom events appear to be more important in the overall assessment process than concern ov er within-week autocorrelation. Since exposure distributions vary over time, samples collected across the full range of procedures, activiti es, and environmental conditions that could influence exposures should be included in developing an exposure assessment strategy. Additional ly, this study demontrates that formal statistical tests to assess sta tionarity should be used with care and should always be supported with visual investigation.