Accuracy and Use of Information
Uploaded by
AlexdorwinaAccuracy and Use of Information
Uploaded by
AlexdorwinaAccuracy and Use of Information
Numerous sources of error are expected for project information. While numerical values are
often reported to the nearest cent or values of equivalent precision, it is rare that the actual
values are so accurately known. Living with some uncertainty is an inescapable situation, and
a prudent manager should have an understanding of the uncertainty in different types of
information and the possibility of drawing misleading conclusions.
Forecast uncertainty also exists in the short term. For example, consider estimates of work
completed. Every project manager is familiar with situations in which the final few bits of
work for a task take an inordinate amount of time. Unforeseen problems, inadequate quality
on already completed work, lack of attention, accidents, or postponing the most difficult work
problems to the end can all contribute to making the final portion of an activity actually
require far more time and effort than expected. The net result is that estimates of the actual
proportion of work completed are often inaccurate.
Some inaccuracy in reports and estimates can arise from conscious choices made by workers,
foremen or managers. If the value of insuring accuracy is thought to be low or non-existent,
then a rational worker will not expend effort or time to gather or to report information
accurately. Many project scheduling systems flounder on exactly this type of non-reporting or
mis-reporting. The original schedule can quickly become extremely misleading without
accurate updating! Only if all parties concerned have specific mandates or incentives to report
accurately will the data be reliable.
Another source of inaccuracy comes from transcription errors of various sorts. Typographical
errors, incorrect measurements from reading equipment, or other recording and calculation
errors may creep into the sets of information which are used in project management. Despite
intensive efforts to check and eliminate such errors, their complete eradication is virtually
impossible.
One method of indicating the relative accuracy of numerical data is to report ranges or
expected deviations of an estimate or measurement. For example, a measurement might be
reported as 198 ft. + 2 ft. There are two common interpretations of these deviations. First, a
range (such as + 2) might be chosen so that the actual value is certain to be within the
indicated range. In the case above, the actual length would be somewhere between 196 and
200 feet with this convention. Alternatively, this deviation might indicate the typical range of
the estimate or measurement. In this case, the example above might imply that there is, say, a
two-thirds chance that the actual length is between 196 and 200.
When the absolute range of a quantity is very large or unknown, the use of a statistical
standard deviation as a measure of uncertainty may be useful. If a quantity is measured n
times resulting is a set of values xi (i = 1,2,...,n), then the average or mean value then the
average or mean value is given by:
(14.1)
The standard deviation can be estimated as the square roots of the sample variance s 2,
i.e. , where:
(14.2)
The standard deviation is a direct indicator of the spread or variability in a measurement,
in the same units as the measurement itself. Higher values of the standard deviation indicate
greater and greater uncertainty about the exact value of the measurement. For the commonly
encountered normal distribution of a random variable, the average value plus or minus one
standard deviation, + , will include about two-thirds of x the actual occurrences. A
related measure of random variability is the coefficient of variation, defined as the ratio of the
standard deviation to the mean:
(14.3)
Thus, a coefficient of variation indicates the variability as a proportion of the expected value.
A coefficient of variation equal to one (c = 1) represents substantial uncertainty, whereas a
value such as c = 0.1 or ten percent indicates much smaller variability.
More generally, even information which is gathered and reported correctly may be interpreted
incorrectly. While the actual information might be correct within the terms of the data
gathering and recording system, it may be quite misleading for managerial purposes. A few
examples can illustrate the problems which may arise in naively interpreting recorded
information without involving any conceptual understanding of how the information is
actually gathered, stored and recorded or how work on the project actually proceeds.
