Scribd
Upload
3 views3 pages

Tut 3

The document outlines the concepts and methodologies related to Multidimensional Scaling (MDS) and the INDSCAL model, focusing on their applications in analyzing dissimilarities among objects and preferences in various contexts. It includes exercises for practical application of MDS, such as finding coordinates of objects based on distance matrices and evaluating similarity among business relationships and consumer preferences. Additionally, it discusses the segmentation of hotel customers based on various attributes using the INDSCAL model.

Uploaded by

Chan Hufflepuff
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
3 views3 pages

Tut 3

The document outlines the concepts and methodologies related to Multidimensional Scaling (MDS) and the INDSCAL model, focusing on their applications in analyzing dissimilarities among objects and preferences in various contexts. It includes exercises for practical application of MDS, such as finding coordinates of objects based on distance matrices and evaluating similarity among business relationships and consumer preferences. Additionally, it discusses the segmentation of hotel customers based on various attributes using the INDSCAL model.

Uploaded by

Chan Hufflepuff
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 3

STAT3613

Tutorial 3 2024/10/08

Review
Multidimensional Scaling (MDS)
 Problem settings:
( )
n objects, = dissimilarities
,1 ≤ < ≤

 Objective of MDS:
For the ith object (i = 1, …, n), try to find a p-dimensional configuration =
,…, such that the calculated distance based on

= ( − ) + ⋯+ −
= ( − ) ( − ). ≤ < ≤
match the same ordering as for .

 Metric MDS
When the interesting problem involves with the distance or the metric proximity
data.
a) Transform the distance matrix S into = ( ) × such that
1 1 1 1
=− + + −
2 2 2 2
b) Find the eigenvalues of B:
| − | = 0, solve for 
c) Find the eigenvectors of B:
( − ) = 0, solve for
d) The p-dimensional coordinates of X:
⋯ 0
= ,…, ⎛ ⋮ ⋱ ⋮ ⎞,
0 ⋯
⎝ ⎠
where > >⋯> and we assume , …, ≈0
e) Recalculate the distance matrix as
̂ =( − ) ( − )

INDSCAL model
 Problem settings:
m subjects
( )
n objects, = dissimilarities for each subject
, ,1 ≤ < ≤ ,1 ≤ ≤

1
 Objective
For the ith object (i = 1, …, n), try to find a p-dimensional configuration =
,…, such that the calculated distance based on

, = ( − )
match the ordering of , .
o Object space
=( ,…, )
o Subject space
=( ,…, ) ,
where
= ,…

MDS for preference data


 MDS unfolding

 Respondent 1: A > B > C > E > D


 Respondent 2: E > A > B > D > C

Dimension II Respondent 1
Respondent 2

A
E

Dimension I

 Metric unfolding
 Consider a singular decomposition S = UVT
o Y = V are the ideal vector the subjects
o X = U are the coordinates of the objects
 In some software, standardized data matrix S (among objects) is used,
 Y = (n1)1/2V, X = (n1)1/2U

2
Exercises
1. Do not use computer for this question. A distance of 3 objects A, B, C is given as
A B C
A 0 1 1
B 1 0 1
C 1 1 0

Apply MDS to find the coordinates of the 3 objects. Calculate the Euclidean
distance matrix based on the result of MDS.

2. The managers were asked to rank the similarity of business-to-business


relationships with customers from 10 selected countries. The ranks of the pairwise
comparison of the 10 countries are given in b2b in TUT301.RData, where 1 =
most similar pair to 45 = most different pair. Conduct a MDS to evaluate the
similarity.
a. Which scale do you suggest?
b. How many dimensions do you suggest?
c. Based on a 2-dimensional solution, plot the perceptual map of the countries.
d. Discuss the similarity of the countries and the dimensions.

3. 99 consumers rated 10 different commercial samples of potato crisps in a range of


1-30 (30 corresponds to the highest acceptability). The data are given in the
"TUT302.csv".
a. Conduct a 2-dimensional metric MDS unfolding model to investigate their
preferences. Comment on the variation explained by the 2 - 5 dimensions.
b. Conduct a 2-dimensional non-metric MDS unfolding model to investigate
their preferences.
i. Comment on the variation explained by the 2 dimensions.
ii. Obtain the coordinates of crisps and the ideal vectors of the consumers
iii. Based on the joint plot, identify the consumers’ preferences on the
crisps.

4. **The purpose of this study is to segment hotel customers based on room rates and
rate fences. 74 customers are asked to compare 16 profiles. The profiles are
constructed based on 3 attributes which are advance requirement (days) and rule
type and refundability. The data are stored in TUT303.RData. The similarity data
are store in similarity and the profiles are stored in profiles. Apply the
INDSCAL model using ratio scale to investigate the segmentation.
a. How many dimension(s) do you suggest?
b. Consider a 2-dimensional solution.
i. Plot the profiles labelled by each attribute. Discuss the dimensions.
ii. Plot the customers’ weightings of the dimensions. Discuss the
segmentation.

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy