본문 바로가기
KERD
논문

빅 데이터와 기계 학습의 시대 심리학 연구 모형의 평가 원칙과 방법

Principles and methods for model assessment in psychological research in the era of big-data and machine learning

이태헌
한국심리학회 (The Korean Psychological Association)|2021.12|한국심리학회지:일반 (Korean Journal of Psychology: General)|제40권 제4호|pp. 389-413|DOI http://dx.doi.org/10.22257/kjp.2021.12.40.4.389

0

조회

0

다운로드

0

인용

0

참고문헌

키워드

cross-validation과적합일반화 오차훈련 오차교차-타당성 입증법편향-분산 균형overfittinggeneralization errortraining errorbias-variance tradeoff
주제분류심리과학

초록

본 논문에서는 계량 심리학 분야에서 지난 수 십 년 동안 꾸준히 논의가 진행되어 왔던 모형 추정과 평가의 원칙을 심리학 연구자들에게 소개하는 것을 목적으로 한다. 계량 심리학 분야에서 진행된 논의의 핵심은 1) 후보 모형들은 참 모형(true model)이 아니라 근사 모형(approximating model)이며, 2) 데이터 크기가 무한히 커지더라도 참 모형과 근사 모형 간 불일치는 사라지는 것은 아니기 때문에, 3) 여러 후보 모형 중 참 모형과의 불일치가 가장 낮은 것으로 추정되는 근사 모형을 선정하는 것이 바람직하다는 것이다. 이러한 모형 선정의 원리는 4차 산업 혁명의 시대, 여러 학문 분야에 걸쳐 그 영역을 확장하고 있는 기계 학습(machine learning) 분야에서 채택하고 있는 모형 평가의 원칙과 동일함을 설명하였다. 즉, 기계 학습 분야에서는 훈련(training) 과정에 노출되지 않았던 새로운 사례에서 보이는 모형의 성능인 일반화 혹은 예측 오차(generalization or prediction error)를 추정함으로써 모형을 선정하는데, 이는 계량 심리학 분야에서 근사모형과 참모형의 불일치 추정량인 총체적 오차(overall discrepancy)를 추정함으로써 모형을 선정해야 한다는 원리와 동일함을 설명하였다. 본 논문의 두 번째 목적은, 이러한 모형 선정의 원칙에 대한 이해를 바탕으로, 현재 심리학 분야에서 주어진 데이터에 대한 “철저한” 분석 관행이 초래하는 과적합(overfitting) 문제와 그 해결 방안을 논의하는 데 있다. 특히, 기계 학습 분야에서 가정 널리 사용되고 있으며, 계량 심리학 분야에서도 오래전부터 논의가 되어온(Mosier, 1951) 교차-타당성 입증법(cross-validation)을 일반화 오차의 추정량이라는 관점에서 소개하고 사용을 당부하였다.

Abstract

The objective of the present article is to explain principles of estimation and assessment for statistical models in psychological research. The principles have indeed been actively discussed over the past few decades in the field of mathematical and quantitative psychology. The essence of the discussion is as follows: 1) candidate models are to be considered not the true model but approximating models, 2) discrepancy between a candidate model and the true model will not disappear even in the population, and therefore 3) it would be best to select the approximating model exhibiting the smallest discrepancy with the true model. The discrepancy between the true model and a candidate model estimated in the sample has been referred to as overall discrepancy in quantitative psychology. In the field of machine learning, models are assessed in light of the extent to which performance of a model is generalizable to the new unseen samples, without being limited to the training samples. In machine learning, a model’s ability to generalize is referred to as the generalization error or prediction error. The present article elucidates the point that the principle of model assessment based on overall discrepancy advocated in quantitative psychology is identical to the model assessment principle based on generalization/prediction error firmly adopted in machine learning. Another objective of the present article is to help readers appreciate the fact that questionable data analytic practices widely tolerated in psychology, such as HARKing (Kerr, 1998) and QRP (Simmons et al., 2011), have been likely causes of the problem known as overfitting in individual studies, which in turn, have collectively resulted in the recent debates over replication crisis in psychology. As a remedy against the questionable practices, this article reintroduces cross-validation methods, whose initial discussion dates back at least to the 1950s in psychology (Mosier, 1951), by couching them in terms of estimators of the generalization/prediction error in the hope of reducing the overfitting problems in psychological research.

참고문헌 (69)

  1. Agler, R. A., "Factors associated with sensitive regression weights : A fungible parameter approach", Behavior research methods, 52(1), pp.207~222 (2020)[DOI]
  2. Akaike, H., "Information theory and an extension of the maximum likelihood principle", 2nd International Symposium on Information Theory (1973), Akademia Kiado
  3. Akaike, H., "A new look at the statistical model identification", IEEE transactions on automatic control, 19(6), pp.716~723 (1974)[DOI]
  4. Allen, D. M., "The relationship between variable selection and data agumentation and a method for prediction", technometrics, 16(1), pp.125~127 (1974)[DOI]
  5. Arlot, S., "A survey of cross-validation procedures for model selection", Statistics surveys, 4, pp.40~79 (2010)[DOI]
  6. Probability Series, "Wadsworth Statistics", pp.1984, Breiman, L.
  7. Bozdogan, H., "Akaike's information criterion and recent developments in information complexity", Journal of mathematical psychology, 44(1), pp.62~91 (2000)[DOI]
  8. Browne, M. W., "Cross-validation methods", Journal of mathematical psychology, 44(1), pp.108~132 (2000)[DOI]
  9. Browne, M. W., "Single sample cross-validation indices for covariance structures", Multivariate behavioral research, 24(4), pp.445~455 (1989)[DOI]
  10. Browne, M. W., "Testing structural equation models", pp.136~162 (1993), Sage
  11. Burnham, K. P., "Model Selection and Inference : A Practical Information-Theoretic Approach" (1998), Springer
  12. Burnham, K. P., "Multimodel inference : understanding AIC and BIC in model selection", Sociological methods & research, 33(2), pp.261~304 (2004)[DOI]
  13. Cattell, R. B., "The scree test for the number of factors", Multivariate behavioral research, 1(2), pp.245~276 (1966)[DOI]
  14. Chapman, B. P., "Statistical learning theory for high dimensional prediction : Application to criterion-keyed scale development", Psychological methods, 21(4), pp.603~ (2016)[DOI]
  15. Chatfield, C., "Model uncertainty, data mining and statistical inference", Journal of the Royal Statistical Society : Series A(Statistics in Society), 158(3), pp.419~444 (1995)[DOI]
  16. Chung, H. Y., "A note on bootstrap model selection criterion", Statistics & probability letters, 26(1), pp.35~41 (1996)[DOI]
  17. Cudeck, R., "Cross-validation of covariance structures", Multivariate Behavioral Research, 18(2), pp.147~167 (1983)[DOI]
  18. Cudeck, R., "Model selection in covariance structures analysis and the"problem"of sample size : A clarification", Psychological Bulletin, 109(3), pp.512~519 (1991)[DOI]
  19. Dempster, A. P., "Maximum likelihood from incomplete data via the EM algorithm", Journal of the Royal Statistical Society : Series B(Methodological), 39(1), pp.1~22 (1977)[DOI]
  20. Enders, C. K., "Assessing the fit of structura equation models with multiply imputed data", Psychological Methods, 23(1), pp.76~93 (2018)[DOI]
  21. Geisser, S., "The predictive sample reuse method with applications", Journal of the American statistical Association, 70(350), pp.320~328 (1975)[DOI]
  22. Hastie, T., "The elements of statistical learning: Data mining, inference, and prediction" (2009), Springer
  23. Hu, L. T., "Cutoff criteria for fit indexes in covariance structure analysis : Conventional criteria versus new alternatives", Structural equation modeling : a multidisciplinary journal, 6(1), pp.1~55 (1999)[DOI]
  24. Hurvich, C. M., "Regression and time series model selection in small samples", Biometrika, 76, pp.297~307 (1989)[DOI]
  25. James, G., "An introduction to statistical learning" (2013), springer
  26. Jones, J. A., "Fungible weights in logistic regression", Psychological Methods, 21(2), pp.241~260 (2016)[DOI]
  27. Kerr, N. L., "HARKing : Hypothesizing after the results are known", Personality and social psychology review, 2(3), pp.196~217 (1998)[DOI]
  28. 김청택, "빅데이터를 이용한 심리학 연구 방법", 한국심리학회지:일반, 38(4), pp.519~548 (2019), 한국심리학회[DOI]
  29. Klein, R. A., "Many Labs 2 : Investigating variation in replicability across samples and settings", Advances in Methods and Practices in Psychological Science, 1(4), pp.443~490 (2018)[DOI]
  30. Klein, R., "Data from investigating variation in replicability: A “many labs” replication project", Journal of Open Psychology Data, 2(1) (2014)[DOI]
  31. Krstajic, D., "Cross-validation pitfalls when selecting and assessing regression and classification models", Journal of cheminformatics, 6(1), pp.1~15 (2014)[DOI]
  32. Kuha, J., "AIC and BIC : Comparisons of assumptions and performance", Sociological methods & research, 33(2), pp.188~229 (2004)[DOI]
  33. Kuhn, M., "Applied predictive modeling" (2013), Springer
  34. Lee, T., "Parameter influence in structural equation modeling", Structural Equation Modeling : A Multidisciplinary Journal, 22(1), pp.102~114 (2015)[DOI]
  35. Lee, T., "Fungible parameter estimates in structural equation modeling", Psychological Methods, 23(1), pp.58~75 (2018)[DOI]
  36. Lee, T., "A comparison of full information maximum likelihood and multiple imputation in structural equation modeling with missing data", Psychological Methods, 26(4), pp.466~485 (2021)[DOI]
  37. Lubke, G. H., "Inference based on the best-fitting model can contribute to the replication crisis : Assessing model selection uncertainty using a bootstrap approach", Structural equation modeling : a multidisciplinary journal, 23(4), pp.479~490 (2016)[DOI]
  38. MacCallum, R. C., "Representing sources of error in the common-factor model : Implications for theory and practice", Psychological Bulletin, 109(3), pp.502~511 (1991)[DOI]
  39. Mallow, C. L., "Some comments on Cp", Technometrics, 28, pp.313~319 (1973)
  40. Miller, P. J., "Finding structure in data using multivariate tree boosting", Psychological Methods, 21(4), pp.583~602 (2016)[DOI]
  41. Mitchell, T. M., "Machine Learning" (1997), McGraw-Hill
  42. Mosier, C. I., "Symposium: The need and means of cross-validation. I. Problems and designs of cross-validation", Educational and Psychological Measurement, 11(1), pp.5~11 (1951)[DOI]
  43. Linhart, H., "Model selection" (1986), Wiley
  44. Myung, I. J., "GUEST EDITORS'INTRODUCTION : special issue on model selection", Journal of mathematical psychology, 44(1), pp.1~2 (2000)[DOI]
  45. Nylund, K. L., "Deciding on the number of classes in latent class analysis and growth mixture modeling : A Monte Carlo simulation study", Structural equation modeling : A multidisciplinary Journal, 14(4), pp.535~569 (2007)[DOI]
  46. Pek, J., "Parameter uncertainty in structural equation models : Confidence sets and fungible estimates", Psychological Methods, 23(4), pp.635~653 (2018)[DOI]
  47. Preacher, K. J., "The problem of model selection uncertainty in structural equation modeling", Psychological methods, 17(1), pp.1~ (2012)
  48. Prendez, J. Y., "Measuring Parameter Uncertainty by Identifying Fungible Estimates in SEM", Structural Equation Modeling : A Multidisciplinary Journal, 26(6), pp.893~904 (2019)[DOI]
  49. Press, W. H., "Numerical Recipes with Source Code CD-ROM 3rd Edition: The Art of Scientific Computing" (2007), Cambridge University Press
  50. Raftery, A. E., "Bayesian Model Selection in Social Research", Sociological Methodology, 25, pp.111~163 (1995)[DOI]
  51. "Putting psychology to the test: Rethinking model evaluation through benchmarking and prediction" (2020)[URL]
  52. Schwarz, G., "Estimating the dimension of a model", The annals of statistics, pp.461~464 (1978)
  53. Shi, D., "Understanding the model size effect on SEM fit indices", Educational and psychological measurement, 79(2), pp.310~334 (2019)[DOI]
  54. Simmons, J. P., "False-positive psychology : Undisclosed flexibility in data collection and analysis allows presenting anything as significant", Psychological science, 22(11), pp.1359~1366 (2011)[DOI]
  55. Shrout, P. E., "Psychology, science, and knowledge construction : Broadening perspectives from the replication crisis", Annual review of psychology, 69, pp.487~510 (2018)[DOI]
  56. Stone, M., "Cross-validatory choice and assessment of statistical predictions", Journal of the royal statistical society : Series B(Methodological), 36(2), pp.111~133 (1974)[DOI]
  57. Stone, M., "An asymptotic equivalence of choice of model by cross-validation and Akaike's criterion", Journal of the Royal Statistical Society : Series B(Methodological), 39(1), pp.44~47 (1977)[DOI]
  58. Takeuchi, K., "Distribution of informational statistics and a criterion of model fitting", Suri-Kagaku (Mathematical Sciences), 153, pp.12~18 (1976)
  59. Tukey, J. W., "Exploratory data analysis" (1977), Addison-Wesley
  60. Varma, S., "Bias in error estimation when using cross-validation for model selection", BMC bioinformatics, 7(1), pp.1~8 (2006)[DOI]
  61. Vrieze, S. I., "Model selection and psychological theory : A discussion of the differences between the Akaike information criterion(AIC)and the Bayesian information criterion(BIC)", Psychological Methods, 17(2), pp.228~243 (2012)[DOI]
  62. Waller, N. G., "Fungible weights in multiple regression", Psychometrika, 73(4), pp.691~703 (2008)[DOI]
  63. Ronald L. Wasserstein, "Moving to a world beyond “p< 0.05”", The American Statistician, 73(sup1), pp.1~19 (2019), Informa UK Limited[DOI]
  64. Wiggins, B. J., "The replication crisis in psychology : An overview for theoretical and philosophical psychology", Journal of Theoretical and Philosophical Psychology, 39(4), pp.202~217 (2019)[DOI]
  65. Wherry, R. J., "IV. Comparison of cross-validation with statistical inference of betas and multiple R from a single sample", Educational and Psychological Measurement, 11(1), pp.23~28 (1951)[DOI]
  66. Wherry, R. J., "Underprediction from overfitting : 45 years of shrinkage", Personnel Psychology, 28(1), pp.1~18 (1975)[DOI]
  67. Yarkoni, T., "Choosing prediction over explanation in psychology : Lessons from machine learning", Perspectives on Psychological Science, 12(6), pp.1100~1122 (2017)[DOI]
  68. Yuan, K. -H., "Outliers, leverage observations, and influꠓential cases in factor analysis : Using robust procedures to minimize their effect", Sociological Methodology, 38, pp.329~368 (2008)[DOI]
  69. Zucchini, W., "An introduction to model selection", Journal of mathematical psychology, 44(1), pp.41~61 (2000)[DOI]

관련 논문AI추천

빅데이터를 이용한 심리학 연구 방법김청택한국심리학회 · 한국심리학회지:일반2019 · 제38권 제4호심리학 연구에서 재현성 위기의 현황과 원인 및 대안 모색에 대한 개관고현석, 김빛나, 최준원한국심리학회 · 한국심리학회지:일반2017 · 제36권 제3호구조방정식 모형의 전반적인 평가 및 효과크기와 연속성에 대한 숙고유 소현, 김수영한국심리학회 · 한국심리학회지:일반2024 · 제43권 제3호구조방정식 모형에서 정규성 가정 위배 시 ML의 대안 탐색신 은경, 김수영한국심리학회 · 한국심리학회지:일반2025 · 제44권 제4호구조방정식모형에서 정규화 기법의 적용과 추정 불안정 문제의 해결: 빈도주의 추정법 중심으로김수영, HE ZINING한국심리학회 · 한국심리학회지:일반2025 · 제44권 제3호컴퓨터 기반 적응적 심리 검사 제작을 위한 문항 선정 알고리즘으로서 Alternating Model Tree의 활용 가능성 탐색윤정한, 이태헌한국심리학회 · 한국심리학회지:일반2021 · 제40권 제4호

이 간행물의 다른 논문 (8건)

4차 산업혁명을 맞는 주체성: 과잉과 대안의 성찰최봉영, 한규석pp. 415-457인공지능 편향식별의 공정성 기준과 완화김효은pp. 459-485사용자 특성이 인공지능 수용성 및 인공지능 기반 제품 사용의도에 미치는 영향김남희, 최종안pp. 487-5094차 산업혁명 시대의 진로상담 및 직업심리학: 역할과 과제신주연, 김태선pp. 511-538컴퓨터 기반 적응적 심리 검사 제작을 위한 문항 선정 알고리즘으로서 Alternating Model Tree의 활용 가능성 탐색이태헌, 윤정한pp. 539-566
원문 보기