AAAR 33rd Annual Conference
October 20 - October 24, 2014
Rosen Shingle Creek
Orlando, Florida, USA
Abstract View
Statistical Comparison of Particle Counts
PATRICK O'SHAUGHNESSY, University of Iowa
Abstract Number: 193 Working Group: Instrumentation and Methods
Abstract Particle counters are now commonly used to provide measurements as part of a study to compare aerosol concentrations in different workplaces and ambient air settings. When two settings are to be compared, a t-test is applied to determine the statistical significance of the difference in mean levels of those areas. Proper application of a t-test relies on the assumption that the observations have been randomly obtained from a population and are independent and normally distributed. However, at the high sampling rates often applied when counting particles (< 1 min) there is an increased probability that subsequent observations are autocorrelated which violates the t-test assumptions. Applying a standard t-test to autocorrelated data inflates the probability of a type-1 error relative to its declared value – an increased chance of false rejections of the null hypothesis. Methods have been developed to compensate for autocorrelation that rely on theory associated with the field of time series analysis which requires substantial experience to apply correctly. The study objective was to develop an alternative method that retains the same analytical structure as the standard t-test so that it can be more universally applied by researchers in the aerosols community. The basis of the method is the adjustment of the standard error of the data time series to compensate for the actual increase in this statistic when data is autocorrelated. Methods employed when performing a standard t-test can then be used to compare, for example, the mean of a data series relative to an exposure limit, or the means of two data series. This method can also be applied, after proper adjustments, to lognormally-distributed data, which is very likely when measuring particle counts in occupational and ambient environments. The method was successfully applied to particle counts made in an occupational setting.