Modeling of construction cost data for probabilistic range estimating limited by small data sets of historical costs /

One of the difficulties in forecasting a future construction

Bibliographic Details
Main Author: Evrenosoglu, Faik Burak
Format: Thesis Book
Language:English
Published: [Place of publication not identified] : [publisher not identified] ; 1996.
Subjects:
Description
Summary:One of the difficulties in forecasting a future construction
cost for a given facility type is the process of modeling
historic cost data as a probability distribution. This
process is to facilitate a range estimating methodology
wherein future construction costs are defined as having a
given probability of not being exceeded. Modeling historic
cost as a probability distribution is particularly difficult
when there is a small data set available to the estimator.
In such cases, more than one type of probability distribution
may statistically satisfy the modeling requirements. This
thesis documents the efforts to investigate the statistical
impact on the final estimate of using different probability
distributions for modeling historic construction cost data.
Cost data of seven 2250 bed unit prison facilities, obtained
from Texas Department of Criminal Justice are used for
analysis. Methods of fitting sample data to different
distribution types are reviewed. The problems associated
with estimating the parameters of triangular distribution
based on historical data are identified and a least squares
minimization procedure to estimate the parameters of
triangular distribution is proposed. Specifically, issues
related to triangular, lognormal, gamma, beta, normal and
uniform distributions are discussed. Based on the
statistical impacts to the final cost estimate, lognormal or
gamma distributions are recommended for modeling construction
cost data.
Item Description:Vita.
"Major subject: Civil Engineering".
Physical Description:xiii, 153 leaves : illustrations ; 28 cm.
Issued also on microfiche from Lange Micrographics.
Bibliography:Includes bibliographical references.