Factor Analysis and Methods of Supplier Selection: Tak K. Mak, Fassil Nebebe
Factor Analysis and Methods of Supplier Selection: Tak K. Mak, Fassil Nebebe
Factor Analysis and Methods of Supplier Selection: Tak K. Mak, Fassil Nebebe
variation. Similarly, determining the local relative At a global level, industrial-wise determination of
weight (within a factor) of importance of a criterion the importance and usefulness of these criteria is in
based on its factor loading or correlation with the general based on surveys of a large number of
factor may also lead to results not necessarily in experts using typically dimensional reduction
line with those based on the experiences of the multivariate methods, such as factor analysis [18, 9,
practitioner or expert. These problems arise if the 10]. On the other hand, there is also an extensive
factor analysis is interpreted as a measurement literature on the methodologies for individual
model (of the underlying factors), as it will ignore managers on assessing suppliers based on a set of
the role of the specific factor (residual error) as identified criteria [3]. These methods include
measurement error when in fact the specific factor approaches that attempt to provide a simple,
may represent characteristics of the criterion that is manageable and systematic procedure for
also important for the decision making. In Section weighting the importance of the selection criteria.
3, we discussed the major issues in these methods These methods are summarized and reviewed in
and suggest a new approach that circumvents these [20]. [25] gives an account of recent research
difficulties. To demonstrate the uses of the activities in a wide variety of methodologies. One
proposed method in general multi-criteria problem, such methodology is the analytical hierarchical
we also discussed an example from recruitment process, which breaks down the weight
data to illustrate the application of the proposed assignments to two or more levels, with smaller
method. tasks involving fewer numbers of criteria at each
level. Recently, the uses of dimensional reduction
2. Literature Review methods, such as factor analysis to create a
hierarchical structure [18, 17, 21] have become
increasingly popular. Some researchers [18] use
Selecting the right supplier has direct influence on
further a second level of factor analysis treating the
operating cost as well as the quality of the product
constructs as “items” and the latent variable that
or services provided by the firm [6, 1, 13, 19]. [11]
explains the common variances as a final “score”
showed empirically the importance of supplier
for comparing suppliers.
selection in influencing business performance and
also identified the major impacting factors that 3. Methodology
contribute to such relationship. The work of [24]
and [12] also gave evidence the importance of Consider the general multi-criteria decision
supplier attitude and participation in building a problem involving the selection of an optimal
long term relationship that benefits both the solution from a number of alternatives based on a
supplier and the firm. Because of the importance of potentially large number of relevant criteria. The
supplier selection in supply chain management, determination of criteria for the selection process is
considerable effort has been expended in the assumed to have already been done based on the
research community to develop analytical methods views and judgement of a subject matter expert(s)
that could facilitate this multi-criteria decision and is not the focus of our attention. Instead, a
process [22]. Criteria for selecting suppliers were systematic approach for evaluating the available
discussed in [5, 15, 26, 16], among others. alternatives based on their scores on these criteria
Considerable emphasis is also placed in the will be proposed, harnessing both the power of
literature on categorizing selection criteria into quantitative analyses and past experience and
major factors or constructs [16, 23, 4, 8]. [27] uses judgement of the decision maker. The steps
multivariate methods to examine the relationships involved are:
among various selection criteria. Recently, there
have been considerable interests in extending the 1. Employing dimensional reduction method,
analysis of supplier selection to include additional typically exploratory factor analysis, to
criteria for supplier selection in green supply chain identify a number of constructs. Unlike
management [9, 10]. previous approaches (also see step 2 below)
that use several individual single factor
In general, selection of supplier is a multi-criteria models, multi-factors in an integrated model
decision problem and therefore appropriate (which allows a criterion to load on more than
weighting of the relative importance and relevance one factor) is permitted, broadening
of these criteria is central to the decision process.
3
Int. J Sup. Chain. Mgt Vol. 5, No. 1, March 2016
substantially its scope of application and the four courses. Thus the emphasis is on the
improving its goodness of fit. selected student’s abilities in quantitative analysis
2. Factor rotation performed to obtain and mastering scientific concepts. The performance
interpretable factors which can later be used to score is purely a function of Factor 1 and Factor 2
check for consistence of importance values only. Conditional on the values of the two factors,
assigned to the selection criteria in step 3 the conditional distributions of the scores of the
below. four subjects carry no further relevant information.
3. With references to a conceptual “performance This is in line with the spirit of factor analysis as a
score function”, the decision maker assigns measurement model where the interest is in
important values to the selection criteria based measuring the underlying constructs and the
on the factors identified in steps 1 and 2, using residual (specific factor) in each of the equation in
a new, easily implemented procedure.
Table 1. Factor model of the correlations
We now elaborate the 3 steps for multi-criteria
decision and explain why some of the traditional Factor Factor Unique Commu-
Subject
methods have some potentially major issues that 1 2 -ness nalities
need to be carefully addressed. Many of the issues Math 0.00 0.95 0.0975 0.9025
involved can be illustrated using a simple, easily
understood example. The proposed method will Physics 0.88 0.40 0.0656 0.9344
then be applied to a supplier selection example
discussed in [14]. Consider the following Chemistry 0.85 0.20 0.2375 0.7625
hypothetical situation where scores on four subjects
- Mathematics, Physics, Chemistry and Biology,
are obtained for a sample of 100 students. Suppose Biology 0.50 0.10 0.7400 0.2600
that the following sample correlation matrix is
observed from the scores of the four subjects. Variation
1.75 1.11
explained
correlations (loadings) with a higher level common Quartimax to obtain factors such that each loads
= Λ +
weights are assigned through the two stages of
rating based on the decision maker’s valuable past
I nit ial Eigenv alues Ex traction Sum s of Squared Loadings Rotation Sum s of Squared Loadings
% of % of % of
Fac tor Total Variance Cumulative % T ot al Varia nce Cu mulativ e % Total Variance Cumulat iv e %
1 7. 504 50. 027 50.027 4.500 30.002 30. 002 5.542 36. 945 36.945
2 2. 061 13. 743 63.770 3.967 26.447 56. 449 2.477 16. 510 53.456
3 1. 468 9.785 73.554 1.615 10.764 67. 213 2.139 14. 263 67.719
4 1. 209 8.061 81.615 1.090 7. 267 74. 480 1.014 6.761 74.480
5 .741 4.943 86.558
6 .484 3.227 89.785
7 .344 2.294 92.079
8 .310 2.068 94.147
9 .260 1.731 95.878
10 .206 1.372 97.250
11 .151 1.006 98.256
12 9. 33E-02 .622 98.878
13 7. 63E-02 .509 99.386
14 5. 77E-02 .384 99.771
15 3. 44E-02 .229 100.00 0
Extraction Method: Maxim um Lik elihood.
Using Kaiser’s rule, four factors with eigenvalues factor. For instance, X14 loads on Factors 3 and 4.
greater than 1 are selected. The scree plot (Figure 1 Thus the present example does not meet the
below) also suggests the same thing. assumption in traditional approaches that each item
measures only a single factor. This assumption,
Scree Plot however, is not needed in our suggested method.
8 Factor 1 loads on X2, X5, X6, X8, X10 to X13
which [2, P434] labelled it as “extroverted
6 personality”. Factor 2 loads on X1, X9 and X15
and may be labelled as “suitability”. Factors 3 and
4 4 load on, respectively, X4, X7, X14 and X3, X14
and can be seen as what [2] called “agreeable
2
personality” and “academic” ability.
Eigenvalue
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Table 3. Final rotated factors of the hiring data
Component Number
Consider now other items not in Tables 4 and 5. value of (.2 + .17)/2 = .185. The complete list of
Take for example, X2 (appearance) in S1. Then C(1) importance values are given in Table 6. Given the
= {X8, X7, X4, X3, X14, X11} and X2 is values of X1 to X15 for any given candidate, these
considered to be less important than X4 but more importance values can be used to compute the final
important than X3 and is assigned an importance score of that candidate.
Table 6. Assigned importance values of the 15 items
Item X1 X2 X3 X4 X5 X6 X7 X8
Importance
0.1 0.185 0.17 0.2 0.1 0.185 0.25 1.0
value
Item X9 X10 X11 X12 X13 X14 X15
Importance
0.8 0.185 0.1 0.125 0.1 0.15 1.0
value
7
Int. J Sup. Chain. Mgt Vol. 5, No. 1, March 2016
Two important points are in order. First, the three level of factor analysis model treating
identified factors are statistically independent so “appropriateness of supplier” as the single
that the relative importance of the three factors underlying factor (and factors identified in the low
cannot be determined by the amount of shared level model as the “variables” in the high level
common variances as in [18]. This is a good factor model). As explained in section 3, this is due
illustration of the potential problem of using high to the fact that the residual errors (specific factors)
8
Int. J Sup. Chain. Mgt Vol. 2, No. 1, March 2013
also contain relevant information in addition to room for input from the decision maker based on
what is contained in the common factor. Second, it the person’s skill and past experience.
selection and assessment and implications art review”, Management Science Letters, 2,
for business performance”, Benchmarking: pp. 1465–1490, 2012
An international journal, 10, pp. 472-489, [26] Weber, C. A., Current, J. R. and Benton, W.
2003 C. “Vendor Selection Criteria and Methods”,
[13] Kannan, V.R. and Tan, K.C., “Just in time, European Journal of Operational Research,
total quality management, and supply chain 50, pp. 2-18, 1991
management: understanding their linkages [27] Wu, M.Y. and Weng, Y.C., “A study of
and impact on business performance”, supplier selection factors for high-tech
Omega, 33, 153–162, 2005 industries in the supply chain”, Total Quality
[14] Kendall, M., Multivariate analysis, London: Management & Business Excellence, 21, pp.
Charles Griffin, 1975
[15] Lehmann, D.R. and O'Shaughnessy, J.,
“Decision Criteria Used in Buying Different
Categories of Products”, Journal of
Purchasing and Materials Management, 18,
pp. 9-14, 1982
[16] Nydick, R. L. and Hill, R.P., “Using the
Analytic Hierarchy Process to Structure the
Supplier Selection Procedure”, International
Journal of Purchasing and Materials
Management, 28, pp. 31-33, 1992