Model Building

1
Model Building The Data Analysis Process

• CHAPTER 12

2
Model as means of information

• Modeling is an organizing process for the
  business researcher and decision maker.
• Model must be able to transform data into
  information for it to be considered useful for
  decision-making.

3

• Modeling is an interactive process that is
  usually a combination of theory and newly
  discovered information.
• Modeling tools can be easily divided into two
  groups
• Discovery and hypothesis.

4
Modeling and Data Management

• Data warehousing.
• The turning of gathered data into useful
  decision-making information.
• Data bases encoded and organized by subject.
• Data warehousing software is capable of
  extracting long stored information.

5
Modeling and Data Management

• Data mining.
• Is a knowledge discovery process that extracts
  previously unknown information from very large
  data bases.
• It is a tool that include analytical techniques
  such as descriptive statistics, graphical plots,
  correlation, linear regression, discriminant
  analysis, factor analysis, and data visualization
  techniques.
• See Safeway example exhibit 12.2. Page 356.

6
Other Integrated Mining Tools

• SAS enterprise miner (http//www.sas.com.
• IBM intelligent miner(http//www.software.ibm.com.
  
• It enables users to recognize and classify
  correlation in their warehoused data by
  automatically performing such analyses as cluster
  analysis, predictive modeling, database
  segmentation, and classification that will help a
  company improve its ability to react to markets.

7
Types of Models

• The two basic types of decision models that can
  be used to transform a complex real-world process
  into a more manageable representation of that
  process are
• The verbal model and the mathematical model

8
Types of Models

• The verbal model.
• It is more easily understood by decision makers.
• It can be readily applied to virtually any
  decision problem.
• However, it could be quite difficult to implement
  because of the possibility of omitting many
  implied variables and relationships that could
  affect the objective of the decision maker.

9
Types of Models

• Mathematical model.
• It can describe complex relationships in a
  precise and concise manner.
• It can be directly used to manipulate variables
  to arrive at outcomes.
• Mathematics can easily be programmed for computer
  solution.
• Mathematical model is deterministic in the sense
  that it is assumed that decision variables are
  known, and uncontrollable variables are excluded
  from consideration because their values are not
  known with certainty.

10
Stochastic Probabilistic Model

• STOCHASTIC model is when the decision variables
  are represented by random processes.
• PROBABILISTIC model is when uncontrollable
  variables represented by random processes are
  incorporated into the model via one or more
  possible forms of probability assignment.

11
Model Classification

• The three common elements of classification are
• Level of aggregation/aggregation.
• Micro model of individual or situational
  behavior of a particular company.
• Macro companies behavior in the environment.

12
Model Classification

• Aggregation.
• Disaggregate model individuals choice.
• Aggregate model choice for all individuals.

13
Model Classification

• Time dimension.
• Dynamic changes in process due to changes in
  the variable specification over time, like a
  moving phenomenon. More costly than static.
• Static variables measured at a specific point
  in time i.E a single snapshot.

14
Model Classification

• Degree of uncertainty.
• Deterministic all decision variable values are
  known within a discrete range i.E they operate
  under conditions of uncertainty. This models do
  not rely on the laws of probability for
  estimating variable values.
• Researcher is willing to assume that demand is
  known.
• It include linear programming, break-even
  analysis, linear curves, critical path network
  analysis and time series analysis.

15

• Stochastic or probabilistic the values of one
  or more decision and/or uncontrollable variables
  are not known with certainty, so some form of
  choice process must be used to estimate the
  values of those variables to be used.
• It is more expensive to use this model because of
  their greater computational requirements and more
  complex solution techniques.

16

• This model include decision theory, queuing
  theory, program evaluation review technique
  network analysis (PERT), Monte Carlo simulation,
  and Markov processes.
• In this situation the choice process used for
  this purpose assumes that variable values occur
  randomly, and a relevant probability distribution
  can be ascertained for each variable.

17
Model Specification Managerial Objectives of
Modeling

• Regression ANOVA.
• The four basic objectives of using models to
  solve particular decision problem are
• Classification a.Ka. Taxonomic models used to
  categorize events, entities to aid decision
  making(see appendix A, p.549).

18
Model Specification Managerial Objectives of
Modeling

• Description and explanation or empirical if it
  contains data for parameter estimation.
• Prediction applied at a more operational level
  as a decision making aid for predicting future
  events.
• Optimization or normative modeling.
• It provide suggestions as to the best course of
  action.