ABSTRACT
This chapter discusses the fit data, the Automatic Tank Management System supplies, and the daily pump readings to a linear model using three different regression techniques. These procedures will be weighted least squares, a form of regression that found its origin in James-Stein regression, and measurement error modeling. The chapter explores the inventory reconciliation data for the common problems that can adversely effect calibration coefficient estimates. Small leaks that seem insignificant but persist over long periods of time can lead to devastating toxic plumes emanating from underground storage tanks located at corner gasoline stations, larger tanks located at refineries, bulk storage facilities. The initial meter calibrations will play an important role when evaluating the regression models used to fit the data. The Environmental Protection Agency “requires that dispensing should be metered to within 5 cubic inches for every 5 gallons of product withdrawn”. The manual calibration method provides great precision in estimating the true calibration of the individual meters.
