Transcript Slide 1

Discriminant Analysis
1.
2.
3.
4.
5.
6.
7.
8.
Introduction
Types of DA
Assumptions
Model representation, data type/sample size
Measurements
Steps to solve DA problems
An numerical example
SPSS commands
(to p2)
(to p3)
(to p4)
(to p5)
(to p6)
(to p10)
(to p11)
(to p16)
1
Discriminant Analysis
•
is a powerful statistical tool used to study
the differences between groups of
objects
•
Here, objects could be
1. an individual person or firms, and
2. classifying them can be based on prior or
posterior factors or characteristics
(to p1)
2
Types of DA
• Two groups
– refer to as two-group
discriminant analysis
– Its dependent variable
is termed as
dichotomous
• Three or more group
– Refer to as multiple
discriminant analysis
(MDA)
– Its corresponding
dependent variables
are termed as
multichotomous
(to p1)
3
Assumptions
1) multivariate normality,
– uses the normal probability
plot approach
– uses the most common
statistical tests are the
3) multicollinearity, among
independent variables
– That is to check independent
variables are not correlated to
each other
calculation of skewness value
2) equal covariance
matrices
– Use covariance to check their
corelations
4) Outliers
– "the observations with a
unique combination of
characteristics identifiable as
distinctly different from the
other observations".
(to p1)
4
Model representation
Data type:
Dependent variables = non-metric format
Indep variables = metric format
(to p1)
Sample size : between 5-20 obs for each independent variables
5
Measurements
1.
Group categorizations
2.
Hit ratio
3.
Discriminating power
(to p7)
(to p8)
(to p9)
(to p1)
6
Group categorizations
(to p6)
7
Hit ratio
• Used to measure the model fitness
• Is a maximum chance criteria
Note: We need to compute this value for our original sample size and then compare
(to p6)
to the value that produced by the SPSS; and computer value should not be less than
the formal value in order to claim the significant of fitness of model
8
Discriminating power
(to p6)
References: refer to “hit ratio” for details
9
Steps to solve DA problems
• Step 1: Assess the assumptions
• Step 2: Estimate the discriminant function(s)
and its (their) significance
• Step 3: Assess the overall fit
(to p1)
10
Example
(to p12)
You can obtain this paper by clicking Discriminant paper from my web site
11
• Objective:
– To discriminate the difference practices between the high and
low performance of firms practicing TQM is ISF
– Use score of overall satisfaction as a mean for discriminating
factor
• Steps:
– Step 1, refer to p 762
(to p13)
– Step 2, refer to p763
(to p14)
– Step 3, refer to p763
(to p15)
– Discussion, you can refer to the “discussion” section
(to p1)
12
Step 1, refer to p 762
(to p12)
13
Step 2, refer to p763
(to p12)
14
Step 3, refer to p763
(to p12)
15
SPSS commands
SPSS Windows
(to p17)
16
SPSS windows
•
•
•
•
•
•
•
•
•
Steps to compute Discriminant Analysis
Step 0
Prior the study of analysis, we need to firstly define a new variable as follows:
- Define “group” and assign a value of either 0, 1, 2 to them, as 0 as
neural
Step 1
Select “Analyze”
Select “Classify”
Select “Discriminant”
click “group variable”
–
•
click “define range”
–
–
•
state its max and min ranges
(this range same as min=1, and max=2 for above case)
click “Independent”
–
–
•
•
and select “group” variable as above
select “variables”
that a group of factors that wish to be clustering
Click option “use stepwise method”
select “Statistics”
Learn from iconic base – Pls
refer to my website
17