Evaluation of Validity and Reliability of Diversity...
Transcript of Evaluation of Validity and Reliability of Diversity...
Evaluation of Validity and Reliability ofDiversity Icebreaker Questionnaire
Tetyana Sydorenko
Ladislaus von Bortkiewicz Chair of StatisticsC.A.S.E. – Center for Applied Statisticsand EconomicsHumboldt–Universität zu Berlinhttp://lvb.wiwi.hu-berlin.dehttp://www.case.hu-berlin.de
Motivation 1-1
Diversity Icebreaker (DI)
� Questionnaire developed by HumanFactors AS, Norway
� Measurement of preferences forcommunication and cooperation
� One page questionnaire available in 19languages
� Easy to understand, takes 10 minutes toanswer
Evaluation of Diversity Icebreaker Questionnaire
Motivation 1-2
Application of Diversity Icebreaker
Main goals:� Self assessment� Teamwork
Fields for Application:� Communication and conflict
management� Diversity Management workshops� Leadership skills development
workshops� Intercultural training
Evaluation of Diversity Icebreaker Questionnaire
Motivation 1-3
Research question
Assessment of DI quality using appropriate statistical methods withrespect to:Validity� degree to which evidence and theory support the
interpretations of test (questionnaire) scores� results of the test (instrument) should fit the construct
Reliability� consistency of the measurement instrument (questionnaire)� each item of the instrument should measure the underlying
construct
Evaluation of Diversity Icebreaker Questionnaire
Outline
1. Motivation X
2. Description of Questionnaire3. Research questions4. Methods and Results5. Conclusions
Evaluation of Diversity Icebreaker Questionnaire
Description of Questionnaire 2-1
Concepts of DI
BLUE� Practical, logical, goal oriented� Focus on detail� Use facts, examples, numbers,
calculations� Structured and well prepared
Evaluation of Diversity Icebreaker Questionnaire
Description of Questionnaire 2-2
Concepts of DI
RED� Personal and enthusiastic� Care for others, like to be in a
group� Focus on social consequences and
community spirit� Show trust and consideration
Evaluation of Diversity Icebreaker Questionnaire
Description of Questionnaire 2-3
Concepts of DI
GREEN� Focus on higher ideas, the
larger picture andconnections
� Positive to change, visionary� Open to creative and
innovative ideas� Set high goals� Unstructured and intuitive
Evaluation of Diversity Icebreaker Questionnaire
Research questions 3-1
Importance of personality and abilitymeasurement
� Assessment of skills and knowledgeI selection of candidates for human resourcesI evaluation of learning process (testing)
� The great need for valid and reliable instruments� Commercializing of instruments e.g. TOEFL
Evaluation of Diversity Icebreaker Questionnaire
Research questions 3-2
Evaluation of DI
� Ipsative formatI variables are not independentI correlations ρ = − 1
k−1 , where k is number of variablesI application of Factor Analysis inplausible
� unclear if the questionnaire is valid and reliable withoutipsative format
� validity of DI model for other populations
SolutionProve with appropriate statistical methods that the DI modelfits data in a non-ipsative format in different samples
Evaluation of Diversity Icebreaker Questionnaire
Methods and Results 4-1
Data
� Transformation of questionnaire� Three data sets in new format
Language Norwegian German English
Sample size n 127 117 59
Female in % 40 50 36
Male in % 59 48 64
Average Age na 22.90 24.50
St. Deviation of Age na 3.23 3.99
Table 1: Overview data-sets
Evaluation of Diversity Icebreaker Questionnaire
Methods and Results 4-2
Reliability Analysis: Cronbach’s α
� consistency of latent construct Y =∑m
i=1 Xi , Xicorresponding items
� α proves whether the variance of the latent construct S2Y is
greater than∑m
j=1 S2Xj
α =m
m − 1
S2
Y −m∑
j=1S2
Xj
S2Y
αst =
m · r̄1 + (m − 1) · r̄
Evaluation of Diversity Icebreaker Questionnaire
Methods and Results 4-3
Cronbach’s α Results
Dimension Norwegian German English
BLUE 0.84 0.79 0.82RED 0.85 0.84 0.83GREEN 0.78 0.66 0.79
Table 2: Cronbach’s αst in data sets
� RED items: increase in Cronbach’s α if item v9 was droppedin Norwegian and German sample and v12 in English sample
� GREEN items: increase in Cronbach’s α if item v27 wasdropped in Norwegian sample, v5 in German and English, v7 inGerman data set
Evaluation of Diversity Icebreaker Questionnaire
Methods and Results 4-4
Explanatory Factor Analysis (EFA) forBLUE dimension
Norwegian German English
Horn 2 2 3Kaiser 3 5 5
Table 3: Number of factors for BLUE dimension
Common factors in all data sets:� Preference for working with numbers v28, v31,� Preference for precise communication and decisions v8, v15,
v18, v35� Preference for being practical minded v20, v37
Problematic items: v1, v4, v11, v23, v25, v40
Evaluation of Diversity Icebreaker Questionnaire
Methods and Results 4-5
EFA for RED dimension
Norwegian German English
Horn 2 2 2Kaiser 4 4 4
Table 4: Number of factors for RED dimension
Common factors in all data sets:� Preference to be in a group: v2, v22, v42� Extraversion: v6, v26, v38� Consideration of other’s feelings: v14, v30, v32, v34
Unstable factor:� Preference for personal communication v9, v12, v16
Evaluation of Diversity Icebreaker Questionnaire
Methods and Results 4-6
EFA for GREEN dimension
Norwegian German English
Horn 4 3 1Kaiser 4 5 6
Table 5: Number of factors for GREEN dimension
Common factor in all data sets:� Creative ideas and solutions v33, v36, v39� Broader perspectives v7, v10, v17
Problematic items: v3, v5, v13, v21, v24, v27, v41, v49
Evaluation of Diversity Icebreaker Questionnaire
Methods and Results 4-7
EFA for the whole item pool
Norwegian German English
Horn 7 6 5Kaiser 14 14 12
Table 6: Number of factors for the whole item’s pool
� RED items loaded on one/two factors in all data sets� mixed factors containing BLUE and GREEN items
Evaluation of Diversity Icebreaker Questionnaire
Methods and Results 4-8
Two parameter Item Response Model (2PL)
P(Yij = 1 | θi ) =exp [αj(θi − δj)]
1 + exp [αj(θi − δj)]
� i = 1, . . . , n observations, j = 1, . . . , p items� θi latent trait of person i� δj difficulty of item j ,� αj discrimination power of item j
Goals:� Evaluation of item’s difficulty parameter and separation power
in BLUE/RED/GREEN sets� Estimation of δj , αi and detection of items which function
badlyEvaluation of Diversity Icebreaker Questionnaire
Methods and Results 4-9
2PL for BLUE dimension (German data)
Figure 1: ICC curves for BLUE items in German sample
Evaluation of Diversity Icebreaker Questionnaire
Methods and Results 4-10
2PL for BLUE dimension (English data)
Figure 2: ICC curves for BLUE items in English sample
Evaluation of Diversity Icebreaker Questionnaire
Methods and Results 4-11
2PL for RED dimension (German data)
Figure 3: ICC curves for RED items in German sample
Evaluation of Diversity Icebreaker Questionnaire
Methods and Results 4-12
2PL for RED dimension (English data)
Figure 4: ICC curves for RED items in English sample
Evaluation of Diversity Icebreaker Questionnaire
Methods and Results 4-13
2PL for GREEN dimension (German data)
Figure 5: ICC curves for GREEN items in German sample
Evaluation of Diversity Icebreaker Questionnaire
Methods and Results 4-14
2PL for GREEN dimension (English data)
Figure 6: ICC curves for GREEN items in English sample
Evaluation of Diversity Icebreaker Questionnaire
Methods and Results 4-15
Multiple Indicator Multiple Cause Model(MIMIC)
Is there any significant native speaker effect obBLUE/RED/GREEN dimensions (estimation of betaj)
P(Yij = 1) = G
{δj + αj
(∑k
βkxik + θi
)}
� xik covariate (e.g. gender, native speaker)� G link function (e.g. logistic function)� At the beginning: Confirmatory Factor Analysis (CFA)� Incorporation of covariate and comparison of model fit
Evaluation of Diversity Icebreaker Questionnaire
Methods and Results 4-16
CFA BLUE dimension
Model RMSEA CFI TLI WRMR
one factor
without v23 0.08 0.58 0.50 1.77without v1, v8, v11,v23 v37 0.07 0.88 0.84 0.94without v1, v8, v11,v23, v35, v37 0.09 0.86 0.80 1.06
two factors
F-Numbers and F-Blue 0.05 0.85 0.83 0.90
Table 7: Model fit of BLUE latent trait in German sample
Evaluation of Diversity Icebreaker Questionnaire
Methods and Results 4-17
CFA RED dimension
RMSEA CFI TLI WRMR
all RED items 0.05 0.94 0.92 0.83without v12 0.05 0.94 0.92 0.83with covariate native speaker 0.05 0.92 0.90 0.89
Table 8: Model fit of RED latent trait in German sample
Evaluation of Diversity Icebreaker Questionnaire
Methods and Results 4-18
CFA GREEN dimension
RMSEA CFI TLI WRMR
all GREEN items without v5 0.04 0.85 0.82 0.88without v3,v5, v7, v10, v27 0.09 0.80 0.72 0.99
two factors 0.07 0.80 0.73 0.92
Table 9: Model fit of GREEN latent trait in German sample
Evaluation of Diversity Icebreaker Questionnaire
Methods and Results 4-19
CFA for whole item pool
RMSEA CFI TLI WRMR
all items 0.05 0.63 0.60 1.16without v3, v5, v6, v8, v27 v35 0.03 0.83 0.82 1.01GREEN by v12 0.03 0.84 0.83 1.01
Table 10: Model fit for three factor model in German sample
Evaluation of Diversity Icebreaker Questionnaire
Conclusions 5-1
Conclusions I
Reliability analysis� the GREEN dimension has the lowest value of Cronbach’s α
EFA for single dimension� Factor structure of RED items could be replicated in different
data sets� Different factor models for BLUE and GREEN dimensions in
different data setsEFA for whole item pool� RED items build single factor/factors� Mixed factors with BLUE and GREEN items in all datasets
Evaluation of Diversity Icebreaker Questionnaire
Conclusions 5-2
Conclusions II
2PL Model� RED items had on general higher separation power� More BLUE and GREEN items with low discrimination
coefficientsCFA and MIMIC for single dimension� Accaptable model fit only for the RED dimension� Covariate native speaker does not have significant influence on
RED latent traitCFA and MIMIC for whole item pool� Three factor model does not hold� Especially BLUE and GREEN items should be elaborated
Evaluation of Diversity Icebreaker Questionnaire
Evaluation of Validity and Reliability ofDiversity Icebreaker Questionnaire
Tetyana Sydorenko
Ladislaus von Bortkiewicz Chair of StatisticsC.A.S.E. – Center for Applied Statisticsand EconomicsHumboldt–Universität zu Berlinhttp://lvb.wiwi.hu-berlin.dehttp://www.case.hu-berlin.de
References 6-1
References
William O. Dunlap and John M. CornwellFactor analysis of ipsative measuresMultivariate Behavioural Research 29(1), 1994
Bjørn EkelundThe application of a model which integrates marketsegmentation and psychological theories to change energyconsumption in householdsHenley Management College, 1997
Bjørn Ekelund and Eva LangvikDiversity Icebreaker. How to Manage Diversity ProcessesHuman Factors Publishing, 2008
Evaluation of Diversity Icebreaker Questionnaire