Discriminant Analysis in Sports Presentation on Chapter 10 Presented by Dr.J.P.Verma MSc (Statistics), PhD, MA(Psychology), Masters(Computer Application) Professor(Statistics) Lakshmibai National Institute of Physical Education, Gwalior, India (Deemed University)
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
To understand group differences and to predict the likelihood that a particular entity will belong to a particular class or group based on independent variables
To classify a subject into one of the two groups on the basis of some independent traits.
- Single dependent variable is dichotomous or multichotomous- Independent variables are numeric
3
Application of Discriminant Analysis
Swimmers or Gymnasts on the basis of anthropometric variables High or Low performer on the basis of skills Junior or Senior category on the basis of the maturity parameters
To identify the characteristics for classifying an individual as
4
Similarity between Discriminant Analysis and Regression Analysis
The only difference is in the nature of dependent variable
Dependent Variable
Categorical
Numeric
Use Discriminant Analysis
Use Regression Analysis
5
This Presentation is based on
Chapter 10 of the book
Sports Research with Analytical Solution Using SPSS
a. Identification of independent variables in the model
- Variables having significant discriminating power in classifying a subject into any of the two groups.
b. Function is developed on the identified independent variables - These identified independent variables are used to develop a
discriminating function.
7
Basics of Discriminant Function Analysis
Discriminating variables(Predictors)
Independent variables which construct a discriminant function
Dependent variable(Criterion variable)
Object of classification on the basis of independent variables needs to be categorical Known as Grouping variable in SPSS
8
Basics of Discriminant Analysis
A latent variable which is constructed as a linear combination of independent variables
where b1,b2 … ,bn are discriminant coefficients,X1,X2,…,Xn are discriminating variables and ‘a’ is a constant.
Discriminant function(canonical root)
Z = a + b1X1 + b2X2 + ... + bnXn
9
Basics of Discriminant Analysis
Known as Confusion matrix, assignment matrix or prediction matrix
Used to assess the efficiency of discriminant analysis.
Shows percentage of existing data points that are correctly classified by the model.
Similar to the R2 (percentage of variation in dependent variable explained by the model).
Classification matrix
10
Basics of Discriminant Analysis
Stepwise method of discriminant analysis
Purpose of Study
Confirmatory Exploratory
Develop DF by entering all independent variables
together
Develop DF by entering all independent variables
stepwise
SPSS command
EnterSPSS command
Stepwise
11
Basics of Discriminant Analysis
Capacity of variable to discriminate the cases into any of the two
groups in the model. Determined by the coefficient of the discriminating variable in the
discriminant function. In SPSS output these coefficients are known as standardized
canonical discriminant function coefficients. Higher the value of the coefficient better is the discriminating
power.
Power of discriminating variables
12
Measures the efficiency of discriminant function in the model. Ranges from 0 to 1 Low value of it (closer to 0) indicates better discriminating power of the
model.
Wilk’s Lambda
Basics of Discriminant Analysis
13
Assumptions in discriminant analysis
a. The predictors are normally distributedb. the variance covariance matrices for the predictors within each of
the groups are equal.
Assumptions
What if the assumptions are not satisfied
a. If normality assumption is violated use logistic regression
b. If variance covariance matrices are not equal then use quadratic discriminant technique.
14
Dependent variable is a true dichotomy. The continuous variable should never be dichotomized for the purpose of applying discriminant analysis.
The groups must be mutually exclusive, with every subject or case belonging to only one group.
All cases must be independent. One should not use correlated data like before-after and matched pairs data etc.
Sample sizes of both groups should not differ to a great extent. If sample sizes are in the ratio 80:20 use logistic regression.
Sample size must be sufficient. As a guidelines there should be at least five to six times as many cases as independent variables.
No independent variables should have a zero variability in either of the groups formed by the dependent variable.
Conditions for Discriminant Analysis
15
Why to use discriminant Analysis
To classify the subjects into groups using a discriminant function;
To test a theory by observing whether cases are classified as predicted; To determine the percent of variance in the dependent variable explained
by the independents;
To assess the relative importance of the independent variables in classifying the dependent variable;
To discard those independent variables which do not have discriminating power in classification.
16
Steps in Discriminant Analysis
First Step
Choose independent variables by using either “Enter independents together” and “Use stepwise method” respectively.
Second Step
Develop the discriminant function model by using the coefficients of independent variables and the value of constant in “Unstandardized canonical discriminant function coefficients” table
The discriminant function shall look like as follows
Z = a +b1X1+b2X2+ …….. + bnXn
17
Steps in Discriminant Analysis
Third Step
Wilks’ lambda is computed for testing the significance of discriminant function developed in the model.
Significant value of chi square indicates that the discrimination between two groups in highly significant.
Significance of the model is tested by using classification matrix provided by the SPSS. Also known as confusion matrix.
High percentage of correct classification indicates the validity of the model.
The level of accuracy shown in the classification matrix may not hold for all future classification of new subjects/cases.
Compute Box M statistic to test the equality of variance covariance matrices in the two groups.
18
Steps in Discriminant Analysis
Fourth Step
“Standardized canonical discriminant function coefficients” table is used to find the relative importance of the variables in the model.
Coefficients in the tables is an indication of power of the variable discriminating the two groups.
Fifth Step
A criterion for classification is made on the basis of the mid point of the mean value of the transformed groups if number of cases are same in both groups. Otherwise take weighted average.
If the value of Z calculated with the above mentioned equation is less than this mid value the subject is classified in one group and if it is more than the mid value, it is classified in second group.
19
Application of Discriminant Analysis
To find the discriminatory power of basketball game performance indicators between players at guards and forward positions.
Purpose
Top performing teams during national championships may be selected as subjects for the study.
Sample
Further, only those players who play at guard and forward positions may be selected from the teams for the study.
- A Prototype
20
Plan of Study
The data may be collected from each player by a trained group of observers on the following parameters
Data Collection
Parameters of study
percent of success of 3 point shots percent of success of free-throw shots percent of success of fast-break number of fouls made by number of fouls made on number of defensive rebounds number of offensive rebounds number of turn-over number of steals number of assists number of interceptions number of minutes played.
21
Plan of Study
The objectives of this study can be detailed as follows
To identify independent variables having significant discriminating power in classifying a basketballer into guard or forward position specialist.
To develop a discriminant model for classifying a player into guard and forward position.
To test the validity of model. To find the percentage of correct classification of subjects in the
groups.
Objectives
22
Discriminant Analysis may be used to solve the problem of player’s discrimination by game position.
Test
Output generated by the SPSS
The objectives of the study can be achieved by using the SPSS output. It provides the following five outputs to fulfill the objectives:
Standardized canonical discriminant function coefficients table; Unstandardized canonical discriminant function coefficients table; Functions at group centroids; The value of Wilks’ lambda and significance of chi-square test; Classification matrix.
Plan of Study- Output generated by the SPSS
23
1. Standardized canonical discriminant function coefficients table
Provides standardized discriminant coefficient of each variables. A variable having larger coefficient indicates more discriminating power. The output can be used to show the relative importance of variables in
developing the discriminant function.
2. Unstandardized canonical discriminant function coefficients table
Output contains the nonstandardized coefficients of the variables selected in the model
Used to build the discriminant function
Plan of Study- Interpretation of the Output generated by the SPSS
24
3. Functions at group centroids
Output provides the mean value of the transformed groups Mid point of these two mean is used for classifying a subject in either
of the two groups
Plan of Study- Interpretation of the Output generated by the SPSS
4. The value of Wilks’ lambda and significance of chi-square test
The value of Wilks’ lambda explains the discriminating power of the model
Significant value of chi-square indicates significance of the model in discriminating between two groups.
5. Classification matrix
The fifth output provides the number of subjects classifying correctly into group.
25
Discriminant Analysis with SPSS
ObjectiveTo develop a discriminant function for classifying an individual into sub-junior or junior category
SampleAnthropometric parameters of 10 sub-junior and 10 junior male basketball players.
Research Issues To test the significance of the developed model To assess the efficiency of classification To find relative importance of independent variables retained in the
model
- An Application in Sports
26
Discriminant Analysis with SPSS
ObjectiveTo develop a discriminant function for classifying an individual into sub-junior or junior category
SampleAnthropometric parameters of 10 sub-junior and 10 junior male basketball players.
Research Issues To test the significance of the developed model To assess the efficiency of classification To find relative importance of independent variables retained in the
model 26
- An Application in Sports
27
To buy the book
Sports Research With Analytical Solutions Using SPSS
and all associated presentations click Here
Complete presentation is available on companion website of the book