%0 Journal Article
	%A Bahaa Khalil and  Taha B. M. J. Ouarda and  André St-Hilaire
	%D 2009
	%J International Journal of Environmental and Ecological Engineering
	%B World Academy of Science, Engineering and Technology
	%I Open Science Index 33, 2009
	%T Reconstitute Information about Discontinued Water Quality Variables in the Nile Delta Monitoring Network Using Two Record Extension Techniques
	%U https://publications.waset.org/pdf/6934
	%V 33
	%X The world economic crises and budget constraints
have caused authorities, especially those in developing countries, to
rationalize water quality monitoring activities. Rationalization
consists of reducing the number of monitoring sites, the number of
samples, and/or the number of water quality variables measured. The
reduction in water quality variables is usually based on correlation. If
two variables exhibit high correlation, it is an indication that some of
the information produced may be redundant. Consequently, one
variable can be discontinued, and the other continues to be measured.
Later, the ordinary least squares (OLS) regression technique is
employed to reconstitute information about discontinued variable by
using the continuously measured one as an explanatory variable. In
this paper, two record extension techniques are employed to
reconstitute information about discontinued water quality variables,
the OLS and the Line of Organic Correlation (LOC). An empirical
experiment is conducted using water quality records from the Nile
Delta water quality monitoring network in Egypt. The record
extension techniques are compared for their ability to predict
different statistical parameters of the discontinued variables. Results
show that the OLS is better at estimating individual water quality
records. However, results indicate an underestimation of the variance
in the extended records. The LOC technique is superior in preserving
characteristics of the entire distribution and avoids underestimation
of the variance. It is concluded from this study that the OLS can be
used for the substitution of missing values, while LOC is preferable
for inferring statements about the probability distribution.
	%P 282 - 290