A STRATEGY FOR PSYCHOMETRIC OPTIMIZATION OF [601687]

A STRATEGY FOR PSYCHOMETRIC OPTIMIZATION OF
INTERPERSONAL DEVIANCE SCALE

CORNELIU-ȘTEFAN LIȚĂ,

Institute of Philosophy and Psychology, Romanian Academy;
Psychology Unit, Romanian Gendarmerie

Abstract

Deviant workplace behaviors directed toward co-workers have negative effect on the
social and financial dimensions of any organization and finally lead to the development of
different organizational psychopathologies. The purpose of the current study is twofold: (1)
to analyze the psychometric properties of the Interpersonal Deviance Scale ( DEVI Scale ),
an assessment tool developed by the psychologists of the Romanian Gendarmerie, and (2)
to test a strategy for improving these characteristics. The 34 behaviors of the scale have
been rated by three samples in two conditions ( two samples of 539 and 635 subjects used a
3-point rating scale and other sample of 231 subjects used a 5-point rating scale ). Three
criteria have been taking into account in data analysis: (a) item fit statistics for the Rasch
model, (b) item reliability coefficients, (c) test information function. The preliminary
results suggested that the scale meets the conditions imposed by the Rasch model, having a
very good reliability (0.96-0.98) and a cutting score of 48. Moreover, the 3-point rating
scale presents smaller values of test information function (TIF=14), while the 5-point rating
scale presents higher values (TIF=25). The conclusion is that DEVI scale could be use in
large-scale organizational diagnostic studies to assess the magnitude of abusive behaviors.

Cuvinte cheie: personal militar , devianță interpersonală, analiză Rasch.

Key words : military personnel, interpersonal deviance, Rasch analysis.
1. INTRODUCTION

The workplace has become a forum for the expression of various behaviors
that have great impact to individuals, organizations, and society. Although some of
these organizational behaviors are socially desirable, another set of behaviors may
be viewed as outside normal conventions of acceptability, improper or even very
dangerous. Almost weekly, there are media revelations of wrongdoing in business,
government, educational, and religious institutions ( corruption, violence, or illegal
activity). These behaviors have been analyzed under different labels, such as

 Institute of Philosophy and Psychology, Romanian Academy, Casa Academiei,
Calea 13 Septembrie nr.13, București, e-mail: [anonimizat]

contraproductive behavior (Mangione & Quinn, 1975), unethical decision behavior
(Hegarty & Sims, 1978), withdrawal (e.g., Gupta & Jenkins, 1980), absenteeism
(e.g., Goodman & Atkins, 1984), procedural or distributive injustice (e.g.,
Sheppard, Lewicki, & Minton, 1992), withholding effort (e.g., Kidwell & Bennett,
1993), workplace deviance (Bennett & Robinson, 2003) and antisocial behavior
(Giacolone și Greenberg, 1997). For instance, workplace deviance has been
defined as voluntary behavior that violates significant organizational norms and, in
so doing, threatens the well-being of the organization or its members, or both
(Robinson & Bennett, 1995).
In general, the workplace is seen as a relatively violence free environment,
where confrontation and conciliatory dialogue occurs as part of the normal milieu.
There are instances when this course of events fails to have a positive outcome and
the work setting might be transformed into a hostile and dangerous environment.
Violence in the workplace refers to those incidents where persons are abused,
threatened or assaulted in circumstances relating to their work, involving explicit
or implicit challenge to their safety. For example, World Health Organization
described violence as a generic term incorporating all types of abuse including
behavior that humiliates, degrades or injures the wellbeing, dignity and worth of an
individual. It is true that violence might be regarded as an individual perceptual
experience influenced by a range of factors including culture, context, environment
and past experience (Elliott, 1197), but these abusive behaviors are sometimes
systemic and ingrained in the organization culture, especially for some kind of
institutions ( health care, police, army ).
Deviant workplace behaviors directed toward co-workers or Interpersonal
Deviance ( such as harassment, intimidation, humiliation, scapegoating,
undermining, sabotage, infighting, lying, verbal threats, malicious rumors ) have
negative effect on the social and psychological dimensions of any organization
and finally lead to the development of different organizational
psychopathologies or pose serious economic threat. For example, the annual costs
of workplace deviance have been estimated to be as high as $4.2 billion for
workplace violence alone (Bensimon, 1994), $40 to $120 billion for theft (Buss,
1993; Camara & Schneider, 1994), and $6 to $200 billion for a wide range of
delinquent organizational behavior (Murphy, 1993).
Despite the prevalence and costs of workplace deviance, the understanding of
workplace deviance remains quite limited, and much empirical research has yet to
be done in different types of institutions. Empirical research may be enhanced by
the availability of a validated measure of workplace deviance and the main purpose
of this study is to produce such an accurate instrument that measure Interpersonal
Deviance.

2. METHOD

2.1. OBJECTIVES

The purpose of the current study is twofold:
a) to analyze the psychometric properties of the Interpersonal Deviance
Scale (DEVI Scale), an assessment tool developed by the psychologists of the
Romanian Gendarmerie, and
b) to test a strategy for improving psychometric characteristics.

2.2. MATERIALS

The Interpersonal Deviance Scale (DEVI Scale) has 34 items which have
been formulated starting from observation of specific behavior exhibited within
different military teams/groups. The items presents behaviors related to type of
tasks or tasks distribution, pressures and intimidation, humiliation and
undermining, indifference and patronizing.

2.3. PARTICIPANTS AND PROCEDURES

The 34 behaviors of the scale have been rated by three samples in two
conditions:
• The first group contains 539 subjects from one military organization. The
purpose was to explore data in order to analyze the psychometric properties,
using a 3-point rating scale.
• The second batch contains 635 subjects from four military organizations.
This sample used the same 3-point rating scale and the purpose was to
validate the model obtained initially.
• The third lot contains 231 subjects from five military organizations. In order
to test an optimization strategy for the psychometric characteristic of the
scale, this sample answered the questionnaire using a 5-point rating scale.

2.4. DATA ANALYSIS

In order to achieve the first objective three methods have been used: (a)
descriptive statistics have been computed to analyze the general characteristics of
DEVI Scale, then (b) a Rasch analysis was used to investigate the psychometric
characteristics, and finally (c) a factor analysis was run to explore the
questionnaire dimensionality.
In order to achieve the second objective three criteria have been taking
into account in data analysis: (a) item fit statistics for the Rasch model, (b) item

reliability coefficients, and (c) test information function.
The analyses were conducted with three software programs: Minitab, SPSS
and Winsteps.
3. RESULTS

The results are presented in three parts. The first part shows the exploratory
approach used to understand the psychometric characteristics. The second part
displays the validation approach used to prove the initial results. The third parts
points out the optimization approach used to improve the scale characteristics.

3.1. EXPLORATORY STUDY

Figure 1 presents some descriptive statistics for DEVI Scale and it can be
noticed that results do not have a normal distribution (AD=58,76; p=0.005) – the
most part of the scores are below 40, showing a low level of deviance. In the same
time, there is an important difference between the mean (40) and the median
(36) – and even between their confidence intervals.

Figure no. 1 – Descriptive statistics for DEVI Scale on exploratory sample

The results of the Rasch analysis are displayed in Table 1 and the items have
adequate values at INFIT and OUTFIT statistics, with a small exception – item 1 has
a value of 2.2. at OUTFIT. All items present positive correlations with the overall
score, ranging from .41 to .70.

Table no. 1
Rasch statistics for the items of DEVI Scale on exploratory sample

From Table 2, it can be seen that the mean of INFIT and OUTFIT is around 1,
with an interval of .79 – 1.46 for INFIT and 0.52 – 2.22 for OUTFIT, therefore we
could assume that data fit the Rasch model.

Table no. 2
Rasch Synthesis for DEVI Scale on exploratory sample
The strategy to conduct factor analysis consisted in the Alpha Factoring
method for the extraction of factor followed by Varimax method for factors
rotation. According to Bartlett tests (BTS=12371, sig.=.001) and Kaiser-Meyer-
Olkin (KMO=0,956) we could estimate that data respect the requirements for
factor analysis.
The outcome of factor analysis, presented in Table 3, reveals five factors that
cover 59% of the variance:
 Factor 1 covers 15% of the variance, contains 9 items and could be labeled
Undermining,
 Factor 2 cover 14% of the variance, contains 8 items and could be labeled
Communication restriction ,
 Factor 3 cover 11% of the variance, contains 7 items and could be labeled
Pressures,
 Factor 4 cover 11% of the variance, contains 6 items and could be labeled
Difficult tasks and critics ,
 Factor 5 cover 11% of the variance, contains 4 items and could be labeled
Humiliation.

3.2. VALIDATION STUDY

Figure 2 presents descriptive statistics for DEVI Scale on the second sample
and it can be noticed that results do not have a normal distribution (AD=37,62;
p=0.005) – the most part of the scores are again below 40, showing a low level of
deviance. In the same time, we still have a big difference between the mean (46)
and the median (41), and also among their confidence intervals.

Table no. 3
Factor analysis of DEVI Scale on exploratory sample

Items Factor 1 Factor 2 Factor 3 Factor 4 Factor 5
1 0.20 0.17 0.08 0.56 0.11
2 0.18 0.20 0.16 0.70 0.08
3 0.40 0.25 0.25 0.57 0.09
4 0.32 0.17 0.24 0.64 0.15
5 0.25 0.21 0.36 0.49 0.20
6 0.10 0.25 0.48 0.43 0.27
7 0.29 0.26 0.46 0.32 0.18
8 0.21 0.21 0.59 0.44 0.23
9 -0.04 0.09 0.43 0.42 0.41
10 0.31 0.18 0.66 0.20 0.07
11 0.32 0.23 0.60 0.13 0.13
12 0.25 0.19 0.55 0.20 0.43
13 0.29 0.34 0.32 0.11 0.54
14 0.52 0.35 0.39 0.16 0.01
15 0.39 0.31 0.45 0.20 0.25
16 0.52 0.36 0.43 0.20 0.16
17 0.66 0.19 0.35 0.21 0.17
18 0.69 0.21 0.23 0.31 0.27
19 0.72 0.21 0.19 0.28 0.23
20 0.66 0.29 0.16 0.24 0.09
21 0.34 0.14 0.16 0.28 0.47
22 0.19 0.35 0.25 0.15 0.57
23 0.27 0.62 0.30 0.21 0.12
24 0.30 0.54 0.27 0.24 0.34
25 0.18 0.68 0.17 0.24 0.17
26 0.28 0.70 0.24 0.24 0.07
27 0.19 0.56 0.12 0.24 0.31
28 0.30 0.47 0.10 0.24 0.33
29 0.39 0.49 0.11 0.25 0.32
30 0.50 0.41 0.22 0.21 0.27
31 0.60 0.31 0.18 0.18 0.23
32 0.43 0.42 0.22 0.20 0.29
33 0.20 0.48 0.08 0.10 0.56
34 0.25 0.58 0.38 0.03 0.26

Figure no 2 – Descriptive statistics for DEVI Scale on validation sample

Table 4 illustrates few statistical indicators that help us to understand the
distribution of scores at ID Scale. Hence, we might assume that a result of 48 might
represents the cutting score for interpersonal deviance problems in a specific
organization. For example, Table no.5 shows the results obtained in 4
organizations and it can be seen that the first institution exceed the cutting score,
the second one is close to the limit, while the other two agencies are far below the
critical point.

Table no. 4
Significant statistics of DEVI Scale

Variable Min Quartile 1 Percentile 33 Median Percentile 66 Quartile 3 Max.
ID Scale 34 35 36 41 48 53 102

Table no. 5
The scores of DEVI Scale in four organizations

Organization 1 2 3 4
N N=189 N=141 N=119 N=186
Mean 51,07 46,36 42,22 43.68

Table no. 6
Rasch statistics for the items of DEVI Scale on validation sample

The results of the Rasch analysis are displayed in Table 6 and all the items have
adequate values (even item 1). From the Rasch synthesis, displayed in Table 7, we
could remark that the mean of INFIT and OUTFIT is around 1, with an interval of
.74 – 1.29 for INFIT and 0.56 – 1.70 for OUTFIT, therefore we could conclude that
data fit the Rasch model.
Tabel no. 7
Rasch Synthesis for DEVI Scale on validation sample

Table no. 8
Factor analysis of DEVI Scale on validation sample

Items Factor 1 Factor 2 Factor 3 Factor 4
1 0.23 0.51 0.25 0.21
2 0.26 0.57 0.27 0.18
3 0.25 0.66 0.28 0.20
4 0.24 0.57 0.26 0.31
5 0.29 0.59 0.31 0.28
6 0.36 0.50 0.34 0.32
7 0.47 0.47 0.19 0.39
8 0.38 0.53 0.16 0.34
9 0.23 0.37 0.16 0.61
10 0.56 0.40 0.22 0.29
11 0.44 0.28 0.19 0.35
12 0.40 0.28 0.31 0.46
13 0.42 0.28 0.29 0.51
14 0.44 0.36 0.33 0.14
15 0.43 0.31 0.27 0.36
16 0.43 0.37 0.54 0.23
17 0.28 0.38 0.54 0.29
18 0.21 0.34 0.60 0.34
19 0.24 0.40 0.64 0.24
20 0.22 0.38 0.54 0.25
21 0.32 0.23 0.23 0.60
22 0.21 0.22 0.41 0.59

23 0.64 0.34 0.26 0.24
24 0.40 0.39 0.35 0.47
25 0.67 0.29 0.18 0.29
26 0.50 0.32 0.39 0.34
27 0.41 0.25 0.29 0.45
28 0.43 0.27 0.35 0.39
29 0.47 0.21 0.37 0.41
30 0.54 0.28 0.51 0.20
31 0.32 0.17 0.53 0.37
32 0.31 0.23 0.57 0.38
33 0.21 0.24 0.38 0.59
34 0.68 0.22 0.33 0.26

The factor analysis was conducted using the same method for the extraction of
factor (Alpha Factoring ) followed by a similar method for rotation ( Varimax). The
tests Kaiser-Meyer-Olkin (KMO=0,976) and Bartlett (BTS=14410, sig.=.001)
confirmed that data are suitable for factor analysis. The results of factor analysis,
displayed in Table 8, distinguish four factors that cover 58% of the variance, as
follows:
 Factor 1 cover 16% of the variance, has 12 items and it might be named
Pressures and communication restriction .
 Factor 2 cover 14% of the variance, has 7 items and it might be named Difficult
tasks and critics .
 Factor 3 cover 14% of the variance, has 7 items and it might be named
Undermining .
 Factor 4 cover 14% of the variance, has 8 items and it might be named
Humiliation

3.3. OPTIMIZATION STUDY

The third study aimed to test the impact of the 5-point rating scale on
psychometric properties.
Figure 3 displays descriptive statistics for DEVI Scale on the third sample and
we can observe that results do not have a normal distribution (AD=6,56; p=0.005) –
the most part of the scores are below 60, showing a low level of deviance. In the
same time, we have a smaller difference between the mean (58) and the median
(56) and their confidence intervals overlap.
In Table 9 we can see that the outcome of the Rasch analysis showed
similar results regarding INFIT (.75 – 1.57), OUTFIT (0.73 – 1.47) and reliability
(.97). However, the only difference was related with test information function
which is higher on the 5-point rating scale (25) than on the 3- point rating scale
(12), as we can notice in Figure 4.

Summary for DEVI Lot
40 50 60 70 80 90
95% Confidence
M
ean

52 54 56 58 60

Anderson- Darling Normality Test
A-Squared 6,56
P-Value < 0,005
M ean 58,104
StDev 12,627
Variance 159,433
Skewness 1,05186
Kurtosis 0,51827
N 231
M inimum 39,000
1st Quartile 48,000
M edian 56,000
3rd Quartile 64,000
M aximum 94,000
95% Confidence Interval for M ean
56,467 59,741
95% Confidence Interval for M edian
53,000 57,882
95% Confidence Interval for StDev
11,571 13,896

Figure no. 3 – Descriptive statistics for DEVI Scale on optimization sample

|MODEL RMSE .10 ADJ. SD .63 SEPARATION 6.26 Item RELIABILITY.98 |
MODEL RMSE .10 ADJ. SD .63 SEPARATION 6.26 Item RELIABILITY .98 |
| S.E. OF Item MEAN = .11 |
+––––––––––––––––––––––––––+
Tabel no. 9
Rasch Synthesis for DEVI Scale on optimization sample
+–––––––––––––––––––––––––– +
| RAW MODEL INFIT OUTFIT |
| SCORE COUNT MEASURE ERROR MNSQ ZSTD MNSQ ZSTD |
|––––––––––––––––––––––––––|
| MEAN 419.9 238.0 .00 .10 1.03 .2 1.00 -.1 |
| S.D. 65.3 .0 .64 .02 .19 1.9 .20 1.9 |
| MAX. 536.0 238.0 1.55 .15 1.57 4.7 1.47 3.8 |
| MIN. 296.0 238.0 -.98 .08 .75 -3.1 .73 -2.8 |
|––––––––––––––––––––––––––|
| REAL RMSE .11 ADJ.SD .63 SEPARATION 5.98 Item RELIABILITY .97 |

Figure no. 4 – Test information function for DEVI Scale on validation and optimization
samples

4. CONCLUSION

The purpose of this paper was to present an organizational diagnosis tool
useful to assess the interpersonal deviance at workplace. The study was target to
certify that DEVI Scale respects the fundamental condition to be considered a
measure of interpersonal deviance.
The results obtain on two large samples confirmed that: (a) the scale meets
the conditions imposed by the Rasch model, (b) the scale has a very good
reliability (0,96-0,98), (c) the cutting score could be set at 48, and (d) the scale
has 4 sub-factors. Furthermore, the preliminary results of an optimization study
suggested that: (a) the distribution of results seems to be better if a 5-point rating
scale is used, (b) both types of rating scales have good reliability (.65 and .96 )
although a 5- point rating scale presents higher values of test information function
(25) than a 3-point rating scale (12).
The main limit of the 3-point rating scale is related with its capacity to
discriminate at lower levels of scores ( a small difference between the minimum and
quartile 1). However, it is natural that most scores lie in the lower part of the
interval because the scale investigates atypical behaviors that should not appear
within a normal workplace. Validation study (N= 635) Optimization study (N=231)
3-point rating 5-point rating

In order to improve the analysis, the future studies might try: (a) to verify
the scale ability to discriminate between groups/organizations that exhibit different
degrees of violence, (b) to analyze the impact of different cutting scores on the
classification of organizational deviance, (c) to initiate a comparative analysis with
the results of other scale that assesses interpersonal deviance, (d) to study the
impact of 5-point rating scale.
The conclusion is that Interpersonal Deviance Scale (DEVI Scale) could be
use in large-scale organizational diagnostic studies to assess the magnitude of
abusive/aggressive behaviors.

REFERENCES

1. BENNETT, R. J., & ROBINSON, S. L., The past, present and future of workplace
deviance research in J. GREENBERG (Ed.), Organizational behavior: The state of
the science (2nd ed.), Mahwah, NJ: Erlbaum, 2003, p. 247–281.
2. BENSIMON, H. F., Crisis and disaster management: Violations in the workplace ,
Training and Development, 28, 1994, p. 27–32.
3. BUSS, D., Ways to curtail employee theft . Nation's Business, 38, p. 36.
4. CAMARA, W. J., & SCHNEIDER, D. L. (1994). Integrity tests: Facts and
unresolved issues . American Psychologist , 1993, 49, p. 112-119.
5. ELLIOTT, P., Violence in health care: what nurse managers need to know . Nursing
Management. 1997, p. 38–42.
6. GIACOLONE, R. A., & GREENBERG, J., Antisocial behavior in organizations .
Thousand Oaks, CA: Sage, 1997.
7. GOODMAN, P., & ATKINS, R., Absenteeism. New approaches to understanding,
measuring and managing employee absence . San Francisco: Jossey-Bass, 1984.
8. GUPTA, N., & JENKINS, G., The structure of withdrawal: Relationships among
estrangement, tardiness, absenteeism and turnover . Springfield, VA: National
Technical Information Service, 1980.
9. HEGARTY, W H, & SIMS, H. P, Jr., Some determinants of unethical decision
behavior: An experiment. Journal of Applied Psychology . 63, 1978, p. 451-457.
10. KIDWELL, R. E., & BENNETT, N., Employee propensity to withhold effort: A
conceptual model to intersect three avenues of research . Academy of Management
Review, 18, 1993, p. 429-456
11. MANGIONE, T. W., & QUINN, R. P., Job satisfaction, counterproductive behavior,
and drug use at work . Journal of Applied Psychology , 60, 1975, p.114-116.
12. MURPHY, K. R., Honesty in the workplace. Belmont, CA: Brooks/Cole, 1993.
13. ROBINSON, S., L., BENNETT, R., J., A typology of deviant workplace behaviors: a
multidimensional scaling study . Academy of Management Journal, 1995.
14. SHEPPARD, B. H., LEWICKI, R. J., & MINTON, J. W., Organizational justice: The
search for fairness in the workplace . New York: Lexington Books, 1992.
15. World Health Organisation, The World Health Report , Geneva, WHO, 1997.

REZUMAT

Comportamentele deviante direcționate către colegii de la locul de muncă au efecte
negative asupra dimensiunilor sociale și financiare ale oricărei organizații și duc în final la
dezvoltarea diferitelor tipuri de psihopatologii organizaționale. Prezentul studiu are două
obiective: (1) să analizeze proprietățile psihometrice ale Scalei de Devianță
Interpersonală (Scala DEVI ), un instrument dezvoltat de psihologii din Jandarmeria
Română și (2) să experimenteze o strategie de optimizare a acelor caracteristici. La cele 34
de comportamente incluse în scala respectivă au răspuns trei loturi de cercetare în două
condiții (două eșantioane de 539 și 635 de subiecți au răspuns folosind o scală cu 3 trepte
de evaluare, iar al treilea lot de 231 de subiecți a folosit o scală cu 5 trepte de evaluare ).
În analiza datelor au fost luate în considerare 3 criterii: (a) coeficienții de potrivire
statistică cu modelul Rasch, (b) coeficienții de fidelitate și (c) funcția informațională a
testului. Rezultatele preliminare au sugerat că scala DEVI întrunește condițiile impuse de
modelul Rasch, având o fidelitate foarte bună (0.96-0.98) și un prag de secționare de 48.
Utilizarea scalei cu 3 trepte de evaluare a dus la o valoare mai mica a funcției
informaționale a testului (TIF=14), în timp ce utilizarea scalei cu 5 trepte de evaluare a
prezentat o valoare mai mare (TIF=25). Concluzia este aceea că Scala DEVI poate fi
utilizată în studii ample de diagnoză organizatională în scopul de a măsura magnitudinea
comportamentelor abuzive la locul de muncă.
Titlu in lb. romana

O startegie de optimizare psihometrică a Scalei de Devianță Interpersonală

Colontitlu in engleza (54 de caractere cu spatii cu tot)

A strategy for psychometric optimization of DEVI Scale