EXPLANATION 63
5. Matrix-data Analysis Method
(1) What is the Matrix-data analysis method?
T h ema t r i x - d a t a a n a l y s i s m e t h o d i s a n a p p r o a c h t o c o m p i l i n g m u l t i d i m e
analyzed into two- or three-dimensional data with the loss of i
minimized.
For example, assuming that you are considering buying a car, many i
consumption, prices, style, livability size, drive system, and perform
judgment criteria. It takes a lot of time to evaluate these items one by one. However, it is more
difficult and troublesome for you to compare many candidate cards and select one because:
• there are too many items to evaluate; and
• the difference among the candidate cars is indefinite.
H o w e v e r , i f c r i t e r i a ( y a r d s t i c k s ) a r e e s t a b l i s h e d
performance of cars and show a clear difference among the candidate cars, you can
which car is best. The matrix-data analysis method is a method for obtaining these yardsticks.
(2) Stages in QC stories at which the affinity diagram method should be used
Problem-solving type Step Issue-achieving type
Selecting a theme 1 Selecting a theme
Planning activities 2 Planning activities
Grasping the current situation 3 Making clear issues
Setting goals 4 Setting goals
◎ Analyzing factors 5 Planning measures
Planning measures 6 Pursuing the optimum measure ○
Carrying out the optimum
Carrying out the measures 7
measure
Confirming the effect of Confirming the effect of
the 8 the
measures optimum measure
Entrenching and Entrenching and standardizing
standardizing 9 the
the measures optimum measure
Reflecting on the results Reflecting on the results
and 10 and
determining future activities determining future activities
◎: Frequently used ○: Used
(3) The reason the matrix-data analysis method is included in the N7 tools
The o ther
s a ix (Affinity
pproaches
d iagram
m Relationethod,
d iagram
m Systematicethod,
d iagram
method, Matrix diagram method,Arrow diagram method, and PDPC method) of the N7 tools are
all based on “language data.” The matrix-data analysis method is the only approach that bases
itself on “numerical data.”
Despite this fact, the matrix-data analysis method is included in the N7 tools f
described below.
The relation diagram and affinity diagram methods are approaches to discovering the true cause
by “structuring and diagramming problems” from confused or intricately intertw
data, and the matrix-data analysis method handling numerical data
(structuring and diagramming problems) with these approache
approaches based on language data and the matrix-data analysis method handling num
64 EXPLANATION
data are called the “The New Seven QC Tools.”
EXAMPLE 65
[Step 1] Compile data into a matrix diagram.
[Step 2] Obtain characteristic values, a contributing ratios, and contributing ratio
rows, and determine the principal components to select.
Characteristic value: The degree of importance of each principal component obtained from the
variance (dispersion) of the principal components.
Contributing ratio: The ratio of each characteristic value to the whole obtained from characteristic
value/total variance (the influence exerted by the first principal components).
Contributing ratio row: The cumulative value of contributing ratios based on which the principal
components to be used can be determined.
<Points>
① In many cases, the principal components to be used are determined based on the
characteristic
value and the Cumulative contributing row as criteria.
• When the characteristic value is 1.0 or over (= information of one or more original
variables is included)
• When the Cumulative contributing row is 80% or over
② The matrix-data analysis method is the same approach as “principal component
analysis”
of multivariate analysis.
③ Since the matrix-data analysis method handle a lot of data, manual calculation is virtually
impossible. In general, analyses are carried out using statistical software. When
performing analyses using statistical software, either “correlation matrix” or “variance-
covariance matrix” must be selected. “Correlation matrix” must be used if the units of
variables are different.
[Step 1] Compile data into a matrix diagram.
Student Mathematics Japanese Social Science Student Mathematics Japanese Social Science
(points) (points) studies (points) (points) (points) studies (points)
66 EXAMPLE
(points) (points)
1 43 44 40 55 11 38 49 38 51
2 62 64 48 62 12 42 52 31 54
3 45 44 21 65 13 40 82 62 58
4 37 71 47 54 14 35 82 48 51
5 50 69 37 61 15 62 47 45 67
6 56 39 37 64 16 45 52 19 56
7 56 51 32 65 17 35 47 39 49
8 42 51 32 62 18 53 33 17 62
9 39 44 46 60 19 51 63 51 63
10 57 64 56 64 20 46 63 39 56
[Step 2] Obtain characteristic values, a contributing ratios, and cumulative
contributing ratios, and determine the principal components to select.
Characteristic Contributing ratio
Principal component Contributing ratio
value row
Principal component 1 2.0198 0.5049 0.5049
Principal component 2 1.4753 0.3689 0.8738
Principal component 3 0.3136 0.0784 0.9522
Principal component 4 0.1913 0.0478 1.0000
• The characteristic value and contributing ratio of the first principal component (principal
component 1) are 2.0198 and 0.5049, respectively, and it accounts for 50.49% of the entire
information.
• The characteristic value and contributing ratio of the second principal component (principal
component 2) are 1.4753 and 0.3689, respectively, and it accounts for 36.89% of the entire
information. The contributing ratio row of 0.8738 means that the first and second principal
components account for 87.38% of the entire information.
• Select the first and second principal components because the characteristic value is 1.0 or over
and the contributing ratio row is 80% or over.
PREPARATION 67
[Step 3] Obtain characteristic vectors, and make a model expression for each
principal component.
Characteristic vector: The coefficient when each principal component is given in an expression
(variable weighting)
<Model expression>
First principal component Z1＝ a1 ×Mathematics＋ b1 ×Japanese＋ c1 ×Social studies＋ d1 ×Science
First principal component Z2＝ a2 ×Mathematics＋ b2 ×Japanese＋ c2 ×Social studies＋ d2 ×Science
First principal component Z3＝ a3 ×Mathematics＋ b3 ×Japanese＋ c3 ×Social studies＋ d3 ×Science
Characteristic vectors
To check the relationships between the principal components and the original variables, make a
graph with the principal components selected in Step 2 as the axis/axes. Make a line graph
when only one principal component is selected. When there are three or more principal
components selected, draw a graph combining them.
Factor loading: The correlation coefficient between each principal component and the original
variable
The value is calculated from the expression characteristicvector x
√characteristicvalue and is between -1.0 and 1.0.
[Step 5] Identify the relationship between the principal components and each
variable from the graph, and name the principal components.
Check the relationship between the principal components and each variable on the graph made
in Step 4. Analyze what the principal components forming the axes mean, and interpret them.
The principal components may not always be named. In this case, do not try to give them an
improper name.
68 EXAMPLE
[Step 3] Obtain characteristic vectors, and make a model expression for each
principal component.
Principal Principal Principal Principal
Variable
component 1 component 2 component 3 component 4
Mathematics 0.5599 0.4273 0.0859 0.7046
Japanese -0.4799 0.5026 0.7190 -0.0112
Social studies -0.3489 0.6420 -0.6821 -0.0289
Science 0.5783 0.3906 0.1019 -0.7089
Model expression for each principal component with the characteristic vector as a coefficient
First principal component Z1 = 0.5599 x mathematics – 0.4799 x Japanese – 0.3489 x social
studies + 0.5783 x science
Second principal component Z2 = 0.4273 x mathematics + 0.5026 x Japanese + 0.6420 x
social studies + 0.3906 x science
[Step 4] Obtain and plot factor loadings on a graph.
Principal Principal Principal Principal
Variable
component 1 component 2 component 3 component 4
Mathematics 0.7958 0.5191 0.0481 0.3082
Japanese -0.6820 0.6105 0.4027 -0.0049
Social studies -0.4958 0.7798 -0.3820 -0.0127
Science 0.8219 0.4744 0.0571 -0.3101
Plot factor loadings with the first and second principal components as the axes.
[Step 5] Identify the relationship between the principal components and each
variable from the graph, and name the principal components.
Name each principal component based on the graph as described below.
• Name the first principal component “Trend in literature and science: Science (+ side),
literature (- side)” because mathematics and science are on the + side and Japanese and
social studies are on the – side.
• Name the second principal component “Overall results” because all variables lie on the +
side.
PREPARATION 69
[Step 3] Calculate the points of the principal components.
Calculate the points of each principal component by substituting the sample variables into the
model expression. The position of the sample at each principal component is expressed in a
numerical value.
Principal component points: The position of the sample relative to each principal component
(score)
The principal component points of each student to the first principal component can be
calculated from the model expression obtained in Step 3.
Z1 = 0.5599 x mathematics – 0.4799 x Japanese – 0.3489 x social studies + 0.5783 x science
《Caution》
Even if the variables are substituted into the model expression, the result does not match the
principal component points shown on the right page because data conversion of the following
two types is executed in the process of the calculation of principal component points.
Conversion ①: Standardizing sample data (average = 0 and standard deviation = 1 for each
variable)
↓
Substitute the values standardized in ① into the model expression for each component.
↓
Conversion ②: Standardizing values calculated from the model expressions (average = 0 and
standard deviation = 1 for each principal component)
…Principal component points shown on the right page
[Step 7] Plot the principal component points on the graph to analyze the sample.
It is not easy to identify the actual situation only from the principal component points. Plot them
on a graph to determine the situation.
The graph shows:
• the position of the sample relative to each principal component; and
• the category of the sample.
70 EXAMPLE
[Step 6] Calculate the principal component points.
Principal Principal Principal Principal
Student
component 1 component 2 component 3 component 4
1 -0.1964 -0.7049 -1.3677 0.5233
2 0.5349 1.4524 0.2702 1.8565
3 1.0477 -0.8745 0.9659 -2.0197
4 -1.3611 0.1267 0.3206 -0.3794
5 0.0181 0.5656 1.6302 -0.0188
6 1.2635 0.0756 -0.9952 0.2461
7 1.1452 0.2771 0.6876 -0.0506
8 0.2814 -0.4708 0.3380 -1.7455
9 -0.1236 -0.2991 -1.8874 -1.7640
10 0.2931 1.7273 -0.5753 0.2785
11 -0.8097 -1.0849 -0.9134 0.8045
12 -0.3294 -0.9646 0.2642 0.6750
13 -1.5063 1.4942 0.0018 -1.1350
14 -1.9735 0.2454 1.1192 0.1291
15 1.3987 1.1004 -0.8565 0.3948
16 0.2080 -1.2599 1.6223 0.6954
17 -1.0688 -1.3426 -1.3263 0.8494
18 1.5401 -1.2435 0.3792 0.4162
19 0.0734 1.1697 -0.2943 -0.5044
20 -0.4354 0.0104 0.6169 0.7485
[Step 7] Plot the principal component points on a graph to analyze the sample.
•The top three students are students
10, 13, and 2, and student 13
shows a strong tendency toward
literature.
• Students 3, 6, 7, 15, and 18 are
more likely to be good on science.
Overall
results
Tendency toward
Literature (- side) Science (+ side)
literature/science
EXAMPLE 71
Theme: Company analysis
Data on affiliated companies is obtained. Conduct a comprehensive analysis of the companies
using the matrix-data analysis method.
Operating Ordinary Current Tangible Current Fixed
Company
profit profit assets fixed assets liabilities liabilities
Company A 0.04 0.02 0.68 0.15 0.44 0.32
Company B 0.06 0.03 0.71 0.15 0.53 0.17
Company C 0.04 0.09 0.75 0.16 0.59 0.19
Company D 0.05 0.02 0.75 0.16 0.68 0.13
Company E 0.09 0.07 0.52 0.11 0.41 0.13
Company F 0.07 0.04 0.58 0.20 0.44 0.17
Company G 0.06 0.03 0.65 0.20 0.49 0.24
《Calculating characteristic values, contributing ratios, and contributing ratio rows》
Characteristic
Principal component Contributing ratio Contributing ratio row
value
Principal component 1 2.708 0.451 0.451
Principal component 2 1.615 0.269 0.720
Principal component 3 0.941 0.157 0.877
Principal component 4 0.680 0.114 0.991
Principal component 5 0.031 0.005 0.996
Principal component 6 0.022 0.004 1.000
Judging from the characteristic value and the contributing ratio row, select principal components 1 and 2.
《Calculating characteristic vectors and making model expressions》
Principal Principal Principal Principal Principal Principal
Variable profit
component 1 component 2 component 3 component 4 component 5 component 6
Operating profit 0.567 -0.061 -0.332 -0.090 -0.212 0.714
Ordinary profit
0.224 -0.414 0.504 0.714 0.013 0.115
assets
Current assets
-0.571 -0.232 0.106 -0.057 -0.734 0.257
assets
Tangible fixed
-0.244 0.343 -0.588 0.688 -0.050 0.020
assets liabilities
Current liabilities -0.437 -0.507 -0.230 -0.065 0.599 0.366
Fixed liabilities -0.227 0.630 0.474 0.012 0.229 0.523
First principal component Z1 = 0.567 x operating profit + 0.224 x ordinary profit – 0.571
x current assets – 0.244 x tangible fixed assets – 0.437
x current liabilities – 0.227 x fixed liabilities
First principal component Z2 = -0.061 x operating profit - 0.414 x ordinary profit – 0.232
x current assets + 0.343 x tangible fixed assets – 0.507
x current liabilities + 0.630 x fixed liabilities
《Calculating and plotting factor loadings on a graph》
Principal Principal Principal Principal Principal Principal
Variable profit
component 1 component 2 component 3 component 4 component 5 component 6
Operating profit 0.933 -0.076 -0.322 -0.074 -0.037 0.107
Ordinary profit
0.368 -0.526 0.489 0.589 0.002 0.017
assets
Current assets
-0.939 -0.294 0.103 -0.047 -0.129 0.038
assets
Tangible fixed
-0.401 0.436 -0.570 0.567 -0.008 0.003
assets liabilities
Current liabilities -0.719 -0.644 -0.223 -0.054 0.105 0.055
Fixed liabilities -0.374 0.800 0.460 0.009 0.040 0.079
With the first principal component set as the horizontal axis and the second principal component
set as the vertical axis, plot the factor loadings.
72 EXAMPLE
Liquidity
Profitability
• The large values relative to the first principal component are operating profit and ordinary profit,
and current assets and the other variables lie on the other side. For this reason, name the first
principal component profitability.
• The large values relative to the second principal component are fixed liabilities and tangible
fixed liabilities, and current liabilities and ordinary profit lie on the other side. So, name the
second principal component liquidity.
《Calculating and plotting principal component points on a graph》
Principal Principal Principal Principal Principal Principal
Company
component 1 component 2 component 3 component 4 component 5 component 6
Company A -0.554 1.420 1.256 -0.853 0.406 -0.588
Company B -0.175 -0.290 -0.206 -0.785 -2.081 0.184
Company C -0.674 -1.062 1.193 1.458 -0.058 -0.078
Company D -0.959 -1.053 -1.00 -0.944 1.076 -0.220
Company E 1.970 -0.598 0.526 -0.503 0.497 0.351
Company F 0.599 0.638 -1.117 0.985 -0.118 -1.463
Company G -0.205 0.945 -0.644 0.642 0.279 1.816
EXAMPLE 73
Plot the principal component points on a graph to evaluate each company.
Liquidity
• Company E is superior to the other companies in profitability.
Profitability
• Companies A, C, and D are low in profitability, but the last two are high in liquidity
