확인적 요인분석과 구조방정식모형분석에서 다차원 척도에 대한 다양한 문항묶음방법 적용연구 = The Empirical Study of Various Item Parceling Methods on Multidimensional Scales in CFA and SEM

국문 초록 (Abstract)

본 연구의 목적은 확인적 요인분석과 구조방정식모형분석에서 다차원 척도에 개별문항방법과 다양한 문항묶음 방법(요인·상관·방사·다차원)을 적용하고, 개별문항방법 결과와 다양한 문항묶음방법 결과를 비교분석함으로서 문항묶음방법의 효과성을 파악하는 것이다. 문항묶음은 측정모형에서 여러 관측변수들을 평균이나 합산으로 묶어서 사용하는 방법론이다. 이를 실시했을 때는 자료비정규성 완화, 모형간명화, 측정오차 감소, 공통분산 증가, 모형적합도 안정성, 모수추정 안정성이라는 장점이 있게 된다. 그러나 문항묶음방법론은 다차원 척도에 적용 하게 될 때에 일차원성가정이라는 논쟁점이 제기되며, 다양한 시뮬레이션 연구를 할 때에 모형적합도 및 모수추 정의 안정성 혹은 편향성이라는 상반된 연구결과가 제기되고 있다. 따라서 본 연구는 문항묶음방법론의 논쟁점 에 대한 제언을 제시하고자, 기존 연구와 차별성을 두는 다음의 연구를 진행한다. 첫째, 다차원(다요인)에 속한 하나의 요인에 대하여 일차원성검정을 한 후에, 일차원성이 검정된 하나의 요인에 대하여 문항묶음방법을 적용 한다. 둘째, 다양한 상황의 가설들을 설정하여 문항묶음방법을 적용한 후에, 문항묶음방법이 어떤 경우에는 효과 적이고 어떤 경우에는 비효과적인지 파악한다. 다양한 상황을 고려한 모형은 단순한 다차원(다요인)의 모형, 위계적인 다차원(다요인)의 모형, 작은 표본에 적용한 모형, 표준도구이지만 과다한 요인과 문항을 포함하고 있는 복잡한 다차원(다요인)의 모형, 그리고 다차원(다요인)의 모형들 간의 구조모형이다. 셋째, 다양한 연구모형에 개별문항과 다양한 문항묶음(요인·상관·방사·다차원)을 적용한 후에, 정규성, 확인적 요인분석의 모형적합도 및수렴타당도, 구조방정식모형의 모형적합도 및 효과크기를 검정한다. 다음은 결과이다. 첫째, 확인적 요인분석 결과이다. 문항묶음방법은 개별문항방법과 비교해서, 정규성과 모형적합도(TLI, CFI, RMSEA) 및 수렴타당도(경로계 수, 개념신뢰도, 분산추출지수)가 훨씬 더 안정적이며 좋은 수치로 추정되었다. 둘째, 구조방정식모형분석 결과이 다. 문항묶음방법은 개별문항방법과 비교하여, 모형적합도와 경로계수(직접효과, 간접효과, 총효과)가 약간 더 안정적이었다. 셋째, 다양한 문항묶음방법론비교 결과이다. 확인적 요인분석에는 방사묶음, 구조방정식모형분석에는 다차원묶음이 효과적이었다. 넷째, 다양한 상황의 가설 결과이다. 문항묶음방법은 작은 표본에 적용했을 경우 에, 요인과 문항이 과도하게 많은 복잡한 모형에 적용했을 경우에, 정규성과 모형적합도 및 경로계수가 매우 안정적으로 되었다. 따라서 작은 표본이면서 개별문항수보다 문항묶음수가 크게 감소하는 경우에는 모수 당 표본의 수가 증가되므로, 추정의 안정성은 더욱 향상될 것이다. 여섯째, 문항묶음방법은 정규성가정이 만족되는 자료 보다는 위배되는 자료에 적용했을 경우에, 적합도 및 모수추정이 현격하게 좋아졌다. 본 연구는 문항묶음방법론에 대한 이론적인 설명 및 경험적인 결과를 토대로 하여, 문항묶음방법이 개별문항방법보다는 효과적임을 검정 하면서, 문항묶음방법의 효과적인 적용을 위한 조언도 제시했다.

다국어 초록 (Multilingual Abstract)

The purpose of this study is to apply the methods of factor·correlation·radial·multidimensional parceling to multidimensional scales in Confirmatory Factor Analysis and Structural Equation Model, and then investigate the possibility of the utility of item parceling method by comparing the results of the individual item method with the results of item parceling method. The item parceling method, defined as the sum or mean scores of individual item scores, have been widely used as observation variable in setting up the measurement model in CFA and SEM. The item parceling method is used in data: (1) that is not optimal(nonnormality); (2) that the sample size is small; (3)that has many items in factor; (4) that has hierarchical and complex factors in scale; and (5) that use the standardized scale which is failed to test CFA. The merits of item parceling method are; (1) the alleviation of non normality of data; (2) the simplicity of model; (3) the diminution of measure error index; (4) the increasing of communality; (5) the diminution of unique variance; (6) the stability of model fit; and (7) the stability of parameter estimates. The dispute of item parceling method is the assumption of unidimension and the results of stability or deviation of parameter estimate and model fit in simulation research. The model of research in this study are as follows. The CFA is testified in the model; (1) which has small factors in scale; (2) which has many hierarchical factors in scale; (3) which has small subject; (4) which has many and complex factors and items in standardized scale, and the SEM is testified in the model which has multidimensional factors. The item parceling methods which are composed of factor·correlation·radial·multidimensional method and the individual item method are applied to the measurement variables in models which are explained. The results are as follows. First, the item parceling method is more effective than individual item method in CFA. For example, the indexes of normality, model fit as χ 2 , TLI, CFI, and RMSEA, parameter estimate, and convergent validity as construct reliability and AVE in item parceling method are more stable. Second, the item parceling method is more effective than individual item method in SEM. For example, the indexes of model fit and parameter estimate in item parceling method are more stable. Third, the most effective item parceling method in CFA is radial parceling method and in CFA is multidimensional parceling method. Forth, the item parceling method is effective as it is applied in the model that has small subject. Fifth, the item parceling method is effective as it is applied in the model that has many, complex, and standardized items. Sixth, the item parceling method is more effective than individual item method in SEM. This study suggests the theory of item parceling method from the point of view in psychological measurement and structure equation, the variable methods of item parceling, the merits of item parceling, and the differences between item parceling method and individual item method from empirical research results. This study makes the suggestions that item parceling method can perform to its potential when; (1) the subject size is small; (2) the psychometric properties of the given data is nonnormality; (3) model is complex, (4) there are violations of the assumptions for SEM(e.g., correlated errors), and that item parceling method should not be used in the measurement model which is based on wrong theory and concept.

목차 (Table of Contents)

• Ⅰ. 서 론
• Ⅱ. 이론적 배경
• Ⅲ. 연구방법
• Ⅳ. 연구결과
• Ⅴ. 요약 및 논의

참고문헌 (Reference)

1 이지현, "문항묶음:원리의 이해와 적용" 한국심리학회 35 (35): 327-353, 2016

2 우종필, "구조방정식모델에서 다차원성 개념의 항목묶음 편향에 대한 연구" 한국경영학회 44 (44): 1131-1147, 2015

3 김수영, "구조방정식 모형의 기본과 확장 : mplus 예제와 함께" 학지사 2016

4 조현철, "구조방정식 모델에서 항목묶음이 인과 모수의 검정과 적합도 평가에 미치는 영향" 한국마케팅과학회 17 (17): 133-153, 2007

5 Little, T. D., "Why the items versus parcels controversy needn’t be one" 18 (18): 285-300, 2013

6 Cattell, P. B., "Validation and intensification of the sixteen personality factor questionnaire" 12 : 205-214, 1956

7 Coffman, D. L., "Using parcels to convert path analysis methods into latent variable models" 40 : 235-259, 2005

8 Thompson, B., "Using item “Testlest”/“Parcels” on confirmatory factor analysis: An example using the PPSDQ-78"

9 Kishton, J. M., "Unidimensional versus domain representative Parceling of questionnaire Items : An empirical example" 54 : 757-765, 1994

10 Little, T. D., "To parcel or not to parcel : Exploring the question, weighting the merits" 92 (92): 151-173, 2002

11 Hau, K-T., "The use of item parcels in structural equation modeling : Non-normal data and small sample sizes" 57 : 327-351, 2004

12 Steenkamp, J. E. M., "The use LISREL in validating marketing constructs" 91 (91): 283-299, 1991

13 Enders, C. K., "The relative performance of full information maximum likelihood estimation for missing data in structural equation models" 8 : 430-457, 2001

14 Cattell, P. B., "The radical parcel double design : A solution to the item versus parcel controversy" 10 (10): 165-179, 1975

15 MacCallum, R. C., "The problem of equivalent models in applications of covariance structure analysis" 114 : 185-199, 1993

16 Kenny, D. A., "The performance of RMSEA in models with small degrees of freedom" 44 : 486-507, 2014

17 Daniel, A. Sass, "The effects of parceling unidimensional sales on structural parameter estimates in structural equation model" 13 (13): 566-586, 2006

18 Bandalos, D. L., "The effects of item parceling on goodness-of-fit and parameter estimate bias in structural equation modeling" 9 (9): 78-102, 2002

19 Afifi, T. D., "The chilling effect in families and the pressure to conceal secrets" 72 : 192-216, 2005

20 Corno, L., "The best-laid plans : Modern conceptions of volition" 22 : 14-22, 1993

21 Browne, M. W., "Testing structural equation models" Sage 1993

22 West, S. G., "Structural equation modeling: Concepts, issues, and applications" Sage. Publications 57-75, 1995

23 Holbert, R. L., "Structural equation modeling in the complementary effects of news media use" 28 : 531-551, 2002

24 Anderson, J. C., "Structural equation modeling in practice : A review and recommended two-step approach" 103 (103): 411-423, 1988

25 Brown, K. S., "Situated cognition and the culture of learning" 18 (18): 32-42, 1989

26 Bentler. P. M., "Significance tests and goodness of fit in the analysis of covariance structures" 88 (88): 588-606, 1980

27 MacCallum, R. C., "Sample size in factor analysis" 4 (4): 84-99, 1999

28 Mayer, J. D., "Relation of an ability measure of emotional intelligence to personality" 79 (79): 306-320, 2002

29 Kline, R. B., "Principles and practice of structural equation modeling" The Guildford Press 2011

30 MacCallum, R. C., "Power analysis and determination of sample size for covariance structure modeling" 1 : 130-149, 1996

31 Rogers, W. M., "Parameter recovery and model fit using multidimensional composites : A comparison of four empirical parceling algorithms" 39 : 379-412, 2004

32 Hair, J. F. Jr., "Multivariate data analysis" Person India 1998

33 Hair, J. F. Jr., "Multivariate data analysis" Person India 2009

34 Gardner, H., "Multiple intelligence: The theory of multiple intelligence" Basic Books 1993

35 Pintrich, P. R., "Motivational and self-regulated learning components of classroom academic performance" 82 : 33-40, 1990

36 William, L. J., "Measurement models for linking latent variables and indicators : A review of human resource management research using parcels" 18 : 233-242, 2008

37 Bass, B. M., "Leadership and performance beyond expectations" Free Press 1985

38 Hall, R. J., "Item parceling strategies in SEM : Investigating the subtle effects of unmodeled secondary constructs" 2 : 233-256, 1999

39 Matsunaga, M., "Item parceling in structural equation modeling : A primer" 2 (2): 260-293, 2008

40 Marsh, H. W., "Is more ever too much? The number of indicators per factor in confirmatory factor analysis" 33 : 181-220, 1998

41 Hagtvet, K. A., "How well do item parcels represent conceptually defined latent constructs? A two-facet approach" 11 (11): 168-193, 2004

42 Muthen, L. K., "How to use a Monte Carlo study to decide on sample size and determine power" 9 : 599-620, 2002

43 Weinstein, C. E., "Handbook of research on teaching" Macmillan 1986

44 Mayer, J. D., "Handbook of intelligence" Cambridge University Press 396-420, 2000

45 Campbell, J. P., "Handbook of industrial and organizational psychology" Rand McNally 185-222, 1976

46 Gardner, H., "Frames of mind: The theory of multiple intelligence" Basic Books 1983

47 Fornell, C., "Evaluating structural equation models with unobservable variables and measurement error" 18 (18): 39-50, 1981

48 Finch, J. F., "Effects of sample size and nonnormality on the estimation of mediated effects in latent variable models" 4 : 87-107, 1997

49 Bolye, G. J., "Does item homogeneity indicate internal consistency or item redundancy in psychometric scales?" 3 : 291-294, 1991

50 Field, A., "Discovering statistics using SPSS statistics" UK Sage Publication Ltd 2009

51 Zimmerman, B. J., "Development of a structured interview for assessing student use of self-regulated learning strategies" 23 : 614-628, 1986

52 Bentler, P. M., "Comparative fit indexes in structural models" 107 (107): 238-248, 1990

53 Raykov, T., "Can there be infinitely many models equivalent to a given covariance structure model?" 8 : 142-149, 2001

54 Efron, B., "Boostrap methods : Another look at the jackknife" 7 (7): 1-26, 1979

55 Bandalos, D. L., "Assessing sources of error in structural equation models : The effects of sample size, reliability, and model misspecification" 4 (4): 177-192, 1997

56 MacCallum, R. C., "Applications of structural equation modeling in psychological research" 51 : 201-226, 2000

57 Gerbing, D. W., "An updated paradigm for scale development incorporating unidimensionality and its assessment" 25 : 186-192, 1988

58 Bandalos, D. L., "Advanced structural equation modeling: New developments and techniques in structural equation modeling" IEA 269-296, 2001

59 Graham, J. W., "Adding missing-data-relevant variables to FILM-based structural equation models" 10 : 80-100, 2003

60 Tucker, L. R., "A reliability coefficient for maximum likelihood factor analysis" 38 : 1-10, 1973

61 Bagozzi, R. P., "A general approach to representing multifaceted personality constructs : Application to state self-esteem" 1 : 45-87, 1994

62 Bogozzi, R. P., "A general approach to representation multifaceted personality constructs : Application to state self-esteem" 1 : 35-67, 1994

63 Bogozzi, R. P., "A general approach for representing constructs in organizational research" 1 : 45-87, 1998

64 Muthen, B., "A comparison of some methodologies for the factor analysis of non normal Likert variables : A note on the Size of the Model" 45 (45): 19-30, 1992

65 Landis, R. S., "A comparison of approaches to forming composite measures in structural equation models" 3 : 186-207, 2000

66 Nasser, F., "A Monte Carolo study investigating the impact of item parceling on measures of fit of the in confirmatory factor analysis" 63 : 729-757, 2003

67 Stephenson, M. T., "A Monte Carlo simulation of observable versus latent variable structural equation modeling techniques" 30 : 332-354, 2003

68 Hughey, J., "16PE-E structure using radical parcel versus items" 107 (107): 107-119, 1982