M. Sharma et al., LINEAR-REGRESSION ANALYSES WITH CENSORED-DATA - ESTIMATION OF PAH WASHOUT RATIOS AND DRY DEPOSITION VELOCITIES TO A SNOW SURFACE, Canadian journal of civil engineering, 22(4), 1995, pp. 819-833
Detection limits of analyzing instruments are the main reason for cens
ored observations of pollutant concentrations. An iterative least squa
res method for regression analyses is developed to suit the doubly cen
sored data commonly observed in environmental engineering. The modifie
d iterative least squares method utilizes the expected values of censo
red observations estimated from the probability density function of do
ubly censored data in a regression process. The modified method is exa
mined for bias in the estimation of the parameters of a linear model,
and in the estimation of the standard deviation of the regression. A m
echanistic model for atmospheric transport and deposition of polycycli
c aromatic hydrocarbons (PAHs) to a snow surface is formulated by util
izing the long-term PAH retention property of deep snowpacks. The modi
fied iterative least squares method is applied to estimate the deposit
ion parameters (dry deposition velocity and washout ratio) for various
PAH species, since some of the PAH deposition levels were below the m
inimum detection limit of the analyzing instrument. The estimated para
meters are examined statistically, and compare favourably with previou
sly reported estimates of these parameters.