RISK AND UNCERTAINTY

Prediction From Limited Data

Estimation of Sales Income Based on Expert Opinion
Mathematical Models For Distribution Curves
Normal Distribution
Tables and Associated Integrals of f(Z)
The Cumulative Curve
 Risk and Uncertainty in Project Cash Flows

• NPV and DCFRR for future projects can never be

predicted absolutely because the cash flow data
for such projects are subject to uncertainty.
• Therefore, when stating predicted values of NPV
and DCFRR for projects, you have to give a
measure of confidence in the predictions.
• For example, there is 90 percent chance that
DCFRR will be greater than 10 percent; 50
percent chance of its being greater than 16
percent and only 10 percent of its being greater
than 20 percent.
 Risk and Uncertainty Contd.
• The probability data provide the desirable
information for the decision to proceed with
project or not.
• Estimation of probability requires the use of
statistics.
• Statistical methods play a leading role in
decision making.
 Prediction From Limited Data
• Prediction of future sales price, sales volumes
etc. are based on very limited amount of data
about past event.
• It is not convenient to use the entire
population of past events even when
available.
• A statistic is a measure, based on limited
information from a sample, that allows
parameter of the population to be estimated
 The Mean Value
• The mean value x of a property x is a statistic based on a sample of n
items.
x ( x1  x2  x3    xn ) / n (3.1)
x is the statistic that corresponds to the population parameter
 is the arithmetic average of all the items in the population
• Not all the x values will be different, hence

 
x   xi f ( xi ) /  f ( xi )  (3.2)
f(xi) is the frequency with which a particular value xi occurs.
 The Mean Value
• It is convenient to divide the frequency of occurrence
by the total number of items.
• In this case, f(xi) becomes the relative frequency of
occurrence of the value xi and ∑f(xi) = 1
• The value of x may be discrete or continuous. The
number of sales of cars (recharge cards) in any one
day must be an integer. If the business sells 4 cars, this
represents all possible values of x in the range 3.5 to
4.5.
• When x represents a continuous variable quantity, it is
convenient to take the total or relative frequency of
occurrence within a given range of x values.
 The Mean Value
• These frequencies can be plotted against the mid
values of x to form a histogram.
• The ordinate is the frequency per unit of width x
bar.
• This makes the area under any bar proportional
to the probability that the value of x will lie in the
given range.
• When the relative frequency is plotted as
ordinate, the sum of the area under the bar is
unity.
 Area Under The Curve
• If x is a continuous variable and the internal
ranges are made smaller and smaller, a
smooth curve will eventually result.
• The area under the curve between x1 and x2
represents the probability that a randomly
selected item will have a value x bar lying in
the range x1 and x2
• This is the information that is desired.
 Data From Past Experience
• Data from experience can be used to generate
frequency distribution curves.
• A company must have an efficient commercial
intelligence system to assess market conditions.
• Accuracy of sales forecasting can be increased by
a careful study of past sales records, price trends,
etc.
• Uncertainty of an estimate increases the further
into the future that the estimate is projected.
 Data From Past Experience
• Estimates of sales income are based on the
opinion of experts.
• Experts estimate maximum, minimum and
most likely (modal) values for a quantity.
• Modal value is not necessarily the mid way
between the minimum and maximum values.
• An expert may be asked to estimate the
probability of occurrence of certain values on
each side of the mode.
 Data From Past Experience
• When experts are questioned separately, the
procedure is known as the Dolphic Method.
• The opinion of each expert is assessed by a
coordinator.
• It is useful to compare the past prediction of
each expert with results obtained in practice.
• The information enables the opinions to be
weighted by the coordinator.
 Data From Past Experience
• The opinion of experts provides the basis for plotting a
frequency or probability distribution curves.
• When the relative frequency is plotted as ordinate, the
area under the curve is unity.
• The area under the curve between two values of a
quantity is the probability that a randomly selected
value will fall in the range between these two values of
quantity.
• These probabilities are mere estimates.
• Their reliability depends on the skill of the forecaster
 Home Assignment
• Get a book on Further Mathematics (SS2/SS3)
• Go through the chapter on Statistics
• Work through the examples on mean and
modal values, standard deviation, plotting of
frequency distribution etc.
• Relate these to what we are doing in Process
economics: sales income, fixed equipment
cost, annual expenditure.
Mathematical Models For Distribution
Curves
• Mathematical models have been developed to fit
various distribution curves.
• Any frequency distribution curve obtained in practice
will exactly fit a curve from any of the mathematical
models.
• Approximation are useful, because practical data are
inherently inaccurate.
• The most common mathematical modes are:
 The binomial distribution
 The Poisson distribution
 The Normal or Gaussian distribution
 The Normal Distribution
• A normal distribution curve is bell shaped.
• The curve obeys the relationship:

f ( x) 

exp  x    / 2 2
2
 (3.3)
 2 0.5
 is the true s tan dard deviation
s0 is the s tan dard deviation from a sample
0.5
0
  xi  x  f xi 
2

s   (3.4)
  f xi  1 
 The Normal Distribution

Equation (3.3) can be written in a compact form

as:
   z2 
f ( z )  exp  
  / 2 0.5 (3.5)
  2 
where z is the s tan dard score

z
x    (3.6)

 The Normal distribution
• The area under the curve of f(z) is unity if the
abscissa extends from minus infinity to plus
infinity.
• The area under the curve between z1 and z2 is
the probability that a randomly selected value
of z will be in the range z1 and z2.
• See also material on Normal Distribution
 Mathematical Models
• An event that will definitely occur has a
probability of unity.
• An event that will definitely not occur has a
probability of zero.
• Equation (3.5) can be integrated between limits
to determine probability that a random value lies
between the selected limits.
• Extended tables of f(z) and associated integrals
are available in standard books (Handbook for
Petroleum Engineers).
 Mathematical Models
• A frequency distribution curve can be used to
plot the cumulative frequency curve.
• This is the curve that is most important in
business decisions.
• It can be plotted from a normal frequency
distribution curve.
• The cumulative curve represents the
probability of a random value z having a
value of say, zi, or less.
 Mathematical Models
• If a property C is a function of several
variables, x1, x2 ………xn
C = φ(x1 , x2, ……….xn) (3.7)
• If each x is normally distributed independent
variable, then:
2 2 2
 C  2  C  2  C  2
    1     2  .............     n (3.8)
2
c
 x1   x2   xn 
 Mathematical Models
• бC is the standard deviation of the variable C
• Б1 , б2 , ……………, бn are the standard
deviations of the variables x1 , x2 , ……xn
• Many distributions occurring in business are
not symmetrical but skewed. The normal
distribution is not a good fit.
• However, when data are based on estimates
of future trends, the accuracy of the normal
distribution approximation is acceptable.
 Mathematical Models
• This is particularly the case as the number of
component variable increases.
• An event must have one of two outcomes.
• It must either occur with a probability P1 or
fail to occur with a probability P2.
• These are exclusive events.
• The probability that something will happen is
unity. P1 + P2 = 1
Example on Probability Calculation