LECTURE 6 Input Modelling

There are four steps in the development of a useful model of input data:

1. Collect data from the real system of interest. This often requires a substantial
time and resource commitment. Unfortunately, in some situations, it is not
possible to collect data.
2. Identify a probability distribution to represent the input process.
3. Choose parameters that determine a specific instance of the distribution of
the family.
4. Evaluate the chosen distribution and the associated parameters for
goodness-of-fit.

LECTURE 6 Input Modelling

Distribution
Identification
Data Collection

Histograms
Selecting a Family
of Distributions
Q-Q Plots

Goodness of Fit
Parameter
Estimation

Chi-Square
KolmogorovSmirnov
p-Values

LECTURE 6 Input Modelling

Data Collection
(See Example 9.1)
1.
2.
3.
4.

A useful expenditure of time is planning.

Try to analyse the data as they are being collected.
Try to combine homogeneous data sets.
Be aware of the possibility of data censoring, in which a quantity of interest is
not observed in its entirety.
5. To determine whether there is a relationship between two variables, build a
scatter diagram.
6. Consider the possibility that a sequence of observations which appear to be
independent may possess autocorrelation.
7. Keep in mind the difference between input data and output or performance
data.

LECTURE 6 Input Modelling

Data Distribution Identification

Examples of Probability Distributions

[See pages 276-277 (page 144 in .pdf) of
Discrete-Event System Simulation. 4th Edition]

1.
2.
3.
4.
5.
6.
7.

Binomial
Negative Binomial
Poisson
Normal
Lognormal
Exponential
Gamma

8. Beta
9. Erlang
10. Weibull
11. Discrete or Continuous Uniform
12. Triangular
13. Empirical

LECTURE 6 Input Modelling

Hypothesis Testing

H0:

= or s =

H1:

or s
> or s >
< or s <

Reject H0 when:

2
02 > ,1
( = 1)

<

Nature of H0

Decision

TRUE

FALSE

Reject H0

Type I Error

Fail to reject
H0

Type II Error

LECTURE 6 Input Modelling

Parameter Estimation
Distribution

Parameter(s)

Suggested Estimator(s)

Poisson

1
=

Exponential

Gamma

1
=

Normal

, 2

2 = 2

2 = 2

Lognormal

, 2
(after taking of the data)

LECTURE 6 Input Modelling

Goodness-of-Fit Tests

1. Chi-Square Tests 2
2. Kolmogorov-Smirnov Goodness-of-Fit Test
3. p-Values and Best Fits

LECTURE 6 Input Modelling

Selecting Input Models without Data

Ways to obtain information about a process if data are not available:

1.
2.
3.
4.

Engineering Data
Expert Option
Physical or Conventional Limitations
The Nature of the Process