Autocorrelation in traditional sampling

Hi gang.

Does autocorrelation come into play in traditional (I.e. full shift) sampling programs?

I feel logically that random selection of work days over a year would be preferable to successive day to day sampling crammed into a week.

Would love to hear your thoughts.

John P

Hello, I translated that book section I wrote in 2004 :slight_smile:

Extracted and translated from Lavoué, J.; Deadman, J.E. (2004) Quantitative Surveys in Occupational Health: Strategies for Exposure Assessment and Data Interpretation. In: Occupational Hygiene Manual - from diagnosis to control of risk factors (in French), pp. 391-437. B. Roberge; J.E. Deadman; M. Legris; L. Menard; M. Baril, Eds. Modulo-Griffon, Mont-Royal (QC)

3.8.1 Autocorrelation

Autocorrelation, or serial correlation, refers to the relationship that may exist between measurements taken close together in time. Thus, it is theoretically plausible that measurements taken one after the other correspond to relatively similar exposure conditions (e.g., production rate or climatic conditions). This implies that measurements taken in this way do not constitute an independent sample and therefore no longer meet the assumptions necessary for statistical inference. Francis et al. have clearly demonstrated by simulation that estimating parameters from correlated measurements leads to a bias in the mean and an underestimation of variability (Francis et al., 1989). Several field studies have also demonstrated the existence of such correlation between short-term measurements (between 5 and 30 min) within a day (Kumagai et al., 1993; Kumagai & Matsunaga, 1994; van der Woord et al., 1999). Depending on the author, measurements should be taken between 30 and 60 minutes apart to avoid the biases described above. In the case of eight-hour FEV1, the results of two studies do not suggest the presence of serial correlation between measurements taken from one day to the next (Francis et al., 1989; George et al., 1995). However, several authors have observed an increase in the estimated standard deviations for certain groups as the time difference between measurements increases, suggesting the presence of autocorrelation (Buringh & Lanting, 1991; Deadman et al., 1996; Symanski & Rappaport, 1994). These mixed results suggest that, whenever possible, a study should not be based exclusively on a measurement campaign consisting of consecutive days.

3.8.2 Stationarity
The notion of temporal correlation refers to the stationarity of the defined population. As we saw earlier, it is important to define the study population correctly if valid estimates are to be obtained. It is also important to ensure that the distribution under study is stationary, i.e. that it does not change over time, e.g. due to the installation of control devices. In this context, Symanski et al. recommend as a first approach to define the study population as the set of exposures experienced by a group of workers performing similar tasks over the course of a year (Symanski et al., 1996). Obviously, any significant change in exposure conditions must lead to a new study aimed at characterizing the new distribution corresponding to the population. In this context, Perkins, Mulhausen, and Damiano recommend the use of control charts to highlight systematic changes in exposure levels (Mulhausen & Diamano, 1998; Perkins, 1997).

Buringh, E., & Lanting, R. (1991). Exposure variability in the workplace: Its implications for the assessment of compliance. American Industrial Hygiene Association Journal, 52(1), 6–13.

Deadman, J. E., Armstrong, B. G., & Thériault, G. P. (1996). Exposure to 60-Hz magnetic and electric fields at a Canadian electric utility. Scandinavian Journal of Work, Environment & Health, 22(6), 415–424.

Francis, M., Selvin, S., Spear, R., & Rappaport, S. M. (1989). The Effect of Autocorrelation on the Estimation of Workers’ Daily Exposures. American Industrial Hygiene Association Journal, 50(1), 37–43.

George, D. K., Flynn, M. R., & Harris, R. L. (1995). Autocorrelation of Interday Exposures at an Automobile Assembly Plant. American Industrial Hygiene Association Journal, 56(12), 1187–1194.

Kumagai, S., & Matsunaga, I. (1994). Approaches for Estimating the Distribution of Short-Term Exposure Concentrations for Different Averaging Times. Annals of Occupational Hygiene, 38(6), 815–825.

Kumagai, S., Matsunaga, I., & Kusaka, Y. (1993). Autocorrelation of Short -Term and Daily Average Exposure Levels in Workplace. American Industrial Hygiene Association Journal, 54(7), 341–350.

Mulhausen, J. R., & Diamano, J. (1998). A Strategy for Assessing and Managing Occupational Exposures: Vol. 2nd editio. AIHA Press.

Perkins, J. L. (1997). Modern Industrial Hygiene vol.1—Recognition and Evaluation of Chemical Agents. Van Nostrand Reinhold.

Symanski, E., Kupper, L. L., Kromhout, H., & Rappaport, S. M. (1996). An investigation of systematic changes in occupational exposure. American Industrial Hygiene Association Journal, 57(8), 724–735.

Symanski, E., & Rappaport, S. M. (1994). An investigation of the dependence of exposure variability on the interval between measurements. Annals of Occupational Hygiene, 38(4), 361–372.

van der Woord, M. P., Kromhout, H., Barregard, L., & Jonsson, P. (1999). Within-day variability of magnetic fields among electric utility workers: Consequences for measurement strategies. American Industrial Hygiene Association Journal , 61(1), 31–38.

Translated with DeepL Translate: The world's most accurate translator (free version)

1 Like

To add to the above, the French regulation, which, similar to EN689, requires up to 9 samples, also requires these samples to be separated into three sampling campaigns separated by months. Definitely an attempt at picking up the true exposure variability occuring within a year.

1 Like

Brilliant! Thank you for your response JĂ©rĂ´me!