�ݺ�ߣ

1

Using Friedman test for creating
comparable group results of non-
parametric innovation competence data
Pasi Porkka, M.Sc.
Jari Jussila, M.Sc.
Anu Suominen, M.Sc.

Tampere University of Technology, Pori Unit

4.11.2012
Industrial Management and Engineering

Outline

• Research on human beings and their behavior
• Specific features of nonnumeric and
nonparametric data
• Studied Innovation competence data
• The Friedman test
• Results
• Discussion


Research on human beings and their
behavior

• Some studied aspects are nonnumeric
• Linguistic: questionnaires, evaluations and
interviews
• can be described in words but not measured
with parameters:
• related to feelings, needs, wishes or relations
between humans


Specific features of nonnumeric and
nonparametric data

• Parametrical = there is an implicit functional relationship between
different answers => comparable
• Problems of nonparametric data:
• comparison of results of two different people difficult
• achieving reliable group results
• traditional statistical methods are mathematically and statistically
hardly ever valid to such data ”Quite strong” + ”Strong” = ?

• When humans evaluate something, the inherent method is linguistic
• linguistic variables are nonparametric by nature and therefore not
comparable
• Yet, numbers are easier to store into computers
• For nonparametric data, statistically valid methods are nonparametric


5

Motivation to study competences

• Studies have shown that IQ over 120 does not anymore
differentiate creative from the less creative
(Csikszentmihalyi 1996)
• Emotional intelligence makes up for 80-90 percent of the
competences that differentiate successful leaders from
average leaders (Goleman, Boyatzis, McKee 2002)
• The emotional intelligence is what makes the difference
between successful and average companies (Kets de
Vries 2006)

4.11.2012

Individual’s innovation competence 6

(iceberg model, Spencer and Spencer
1993; Kets de Vries 2001)

Skills (Spencer and Spencer 1993)
Knowledge (Spencer and Spencer 1993)

Self-concept, Attitudes (Spencer
and Spencer 1993)
Traits (Spencer and Spencer 1993)
Motives (Spencer and Spencer 1993)
Emotions (Kets de Vries 2001)
Defenses (Kets de Vries 2001)

4.11.2012

Self-evaluation

• Several statements to each competence (totally
77)
• Linguistic statements evaluated with linguistic
answers, which are transformed into numbers.
• Values are given to current and future state
• Future – current = creative tension (by Senge)
• Answers to statements related to one
competence are summed (with fuzzy logic)
together to gain a single value for a competence
• Each respondent has his/hers own scale of
degree → answers of different respondents are
not comparable


8

Self-evaluation as method

SELF-EVALUATION with linguistic statements with nominal scale

I use theories and models for
s a ll e w s a s w oll a ht o b n oit a zi n a g r o si h T
d e zili vi c a ni et a b e d ot el p o e p s e g a r u o c n e
explaining complicated issues
.r e n n a m

1
e e r g a yl g n o rt S

e er g A
for both
Current: 0,755
current and Saved to database
Target: 0,865
e e r g a si D future states in numeric form Creat.tens: 0,110

e e r g a si d yl g n o rt S
0 Notice! Data is still nonparametric
t e gr a T t n err u C
t e gr a T t n err u C

Industrial Management and Engineering 4.11.2012

9

Self-evaluation data

The data collected with linguistic variables is by nature
weakest in the statistical sense.
1. Measurable only with nominal scale
2. Personal scale of degree (results of different answerers
not comparable)

Values are used merely as means of separating the
properties of elements into different classes.
With nominal scale, traditional statistical methods (sums,
correlations, etc.) are not applicable.
• Own scale of degree and variables are related→ answers
of one respondent are comparable
• Answers of one respondent can be ranked
4.11.2012

The Friedman test: example

Actual numeric values Variables

Respondent Strength of Strength of
father mother
A 0,2 0,1
B 0,3 0,3
C 0,2 0,1
D 0,3 0,2
E 0,3 0,9
SUMS 1,3 1,6


The Friedman test: example

Numeric values transformed into rankings

Respondent Strength of Strength of
father mother
A 2 1
B 1,5 1,5
C 2 1
D 2 1
E 1 2
SUMS 8,5 6,5


Benefits for using Friedman test

• Removes the personal scale of degrees, since actual
given values are turned into rankings
• Sums the rankings - statistically valid group results
• Results comparable to other groups (companies)
• Possible to calculate the minimum difference value,
within which the values are considered equal → great
help in analysis
• Minimum difference depends on amount of
respondents, the distribution of answers → is different
in every group and has to be calculated separately


Averages of creative tensions of two different companies
A n=12 B n=10
Averages of creative tensions of two capacity
0,26 Self-control 0,31 Absorptive
different
companies
0,25 Absorptive capacity 0,28 Professional and technical expertise
0,23 Change orientation 0,27 Self-confidence
0,22 Analytical thinking 0,27 Intuitive thinking
0,21 Conceptual thinking 0,26 Understanding others
0,20 Understanding others 0,24 Analytical thinking
0,17 Self-development 0,22 Communication
0,16 Flexibility 0,21 Accurate self-assessment
0,16 Stress tolerance 0,20 Conceptual thinking
0,15 Professional and technical expertise 0,20 Self-control
0,15 Trustworthiness 0,19 Self-development
0,13 Intuitive thinking 0,19 Flexibility
0,13 Conflict management 0,18 Stress tolerance
0,12 Seeking information 0,15 Relationship building
0,11 Relationship building 0,14 Initiative
0,11 Achievement orientation 0,13 Leveraging diversity
0,11 Accurate self-assessment 0,12 Imagination
0,10 Independence 0,11 Seeking information
0,10 Teamwork and cooperation 0,11 Change orientation
0,09 Self-confidence 0,09 Conflict management
0,08 Divergent thinking 0,09 Teamwork and cooperation
0,08 Responsibility 0,08 Trustworthiness
0,08 Initiative 0,07 Achievement orientation
0,07 Imagination 0,06 Responsibility
0,06 Leveraging diversity 0,05 Divergent thinking
0,04 Communication 0,04 Independence
0,02 Risk orientation 0,00 Risk orientation
3,56 Management and Engineering
Industrial SUM 4,25 SUM

Rankings of creative tensions of two different companies

A n=12, α=0.05, min.diff. 5,71 B n=10, α=0.05, min.diff 5,79
21,83 Self-control 21,70 Absorptive capacity
20,83 Absorptive capacity 21,20 Intuitive thinking
20,25 Change orientation 21,10 Professional and technical expertise
19,17 Conceptual thinking 19,50 Self-confidence
18,83 Analytical thinking 19,30 Understanding others
18,08 Understanding others 18,30 Communication
17,50 Stress tolerance 18,10 Analytical thinking
16,63 Flexibility 17,80 Accurate self-assessment
15,67 Trustworthiness 16,80 Flexibility
15,67 Self-development 16,70 Self-development
14,92 Professional and technical expertise 16,40 Self-control
14,83 Conflict management 16,05 Stress tolerance
14,63 Intuitive thinking 16,00 Conceptual thinking
13,00 Relationship building 14,40 Relationship building
12,29 Seeking information 13,90 Initiative
12,13 Teamwork and cooperation 11,80 Leveraging diversity
11,79 Achievement orientation 11,20 Conflict management
11,63 Accurate self-assessment 11,15 Seeking information
11,46 Self-confidence 10,85 Change orientation
11,29 Divergent thinking 10,70 Imagination
11,17 Initiative 10,15 Responsibility
10,67 Imagination 9,00 Teamwork and cooperation
10,00 Independence 8,30 Achievement orientation
9,88 Responsibility 8,15 Trustworthiness
8,83 Leveraging diversity 7,90 Divergent thinking
7,79 Risk orientation 7,05 Independence
7,25 Communication 4,50 Risk orientation
378,00 SUM
Industrial Management and Engineering 378,00 SUM

Results

Out of the total number of 27 competencies, in both organizations there
are
• 7 corresponding competencies clustered with the high ranked
creative tensions
• 6 corresponding competencies clustered with the low ranked creative
tensions
• The further analysis of the meaning of the similarities in the clusters
remains to be done in the future.

• Significance of the Friedmann test:
1. allows this further analysis to be carried out in the first place.
=> Otherwise sheer speculation and conjecture
2. Friedman test can be carried out to a rather small group of
respondents; however, naturally the group results more reliable the
greater the respondent group is.


Conclusion

• Group results of linguistic data produced with self-
evaluation, although raising questions of invalidity by its
nature can be analyzed with statistically valid method of
Friedman test in order to compare those group results

• However, the further analysis of the gathered results, the
meaning of the clustered rankings, still have to carried out
by a human


Discussion

Why to use the Friedman test?
• Right (statistically) results in analysis with ALL
possible answers
• Minimum difference great help in analysis
• Results comparable with other groups

Why NOT to use?
• Almost the same results as with the (old) summing
style
• Have to learn new method


18

Thank You for your attention

4.11.2012

19

Contact details

Porkka, Pasi
Tampere University of Technology/ Pori
E-mail: pasi.porkka(at)tut.fi

Jussila, Jari
E-mail: jari.j.jussila(at)tut.fi

Suominen, Anu
E-mail: anu.suominen(at)tut.fi

4.11.2012

�ݺ�ߣ

Using Friedman Test For Creating Comparable Group Results Of Non Parametric Innovation Competence Data

More Related Content

Using Friedman Test For Creating Comparable Group Results Of Non Parametric Innovation Competence Data