Evaluating sensitivity to classification uncertainty in latent subgroup effect analyses
Tóm tắt
Increasing attention is being given to assessing treatment effect heterogeneity among individuals belonging to qualitatively different latent subgroups. Inference routinely proceeds by first partitioning the individuals into subgroups, then estimating the subgroup-specific average treatment effects. However, because the subgroups are only latently associated with the observed variables, the actual individual subgroup memberships are rarely known with certainty in practice and thus have to be imputed. Ignoring the uncertainty in the imputed memberships precludes misclassification errors, potentially leading to biased results and incorrect conclusions. We propose a strategy for assessing the sensitivity of inference to classification uncertainty when using such classify-analyze approaches for subgroup effect analyses. We exploit each individual’s typically nonzero predictive or posterior subgroup membership probabilities to gauge the stability of the resultant subgroup-specific average causal effects estimates over different, carefully selected subsets of the individuals. Because the membership probabilities are subject to sampling variability, we propose Monte Carlo confidence intervals that explicitly acknowledge the imprecision in the estimated subgroup memberships via perturbations using a parametric bootstrap. The proposal is widely applicable and avoids stringent causal or structural assumptions that existing bias-adjustment or bias-correction methods rely on. Using two different publicly available real-world datasets, we illustrate how the proposed strategy supplements existing latent subgroup effect analyses to shed light on the potential impact of classification uncertainty on inference. First, individuals are partitioned into latent subgroups based on their medical and health history. Then within each fixed latent subgroup, the average treatment effect is assessed using an augmented inverse propensity score weighted estimator. Finally, utilizing the proposed sensitivity analysis reveals different subgroup-specific effects that are mostly insensitive to potential misclassification. Our proposed sensitivity analysis is straightforward to implement, provides both graphical and numerical summaries, and readily permits assessing the sensitivity of any machine learning-based causal effect estimator to classification uncertainty. We recommend making such sensitivity analyses more routine in latent subgroup effect analyses.
Tài liệu tham khảo
Ferreira JP, Duarte K, McMurray JJV, Pitt B, van Veldhuisen DJ, Vincent J, Ahmad T, Tromp J, Rossignol P, Zannad F. Data-driven approach to identify subgroups of heart failure with reduced ejection fraction patients with different prognoses and aldosterone antagonist response patterns. Circ Heart Fail. 2018;11(7):004926. https://doi.org/10.1161/CIRCHEARTFAILURE.118.004926.
Kim HJ, Lu B, Nehus EJ, Kim M-O. Estimating heterogeneous treatment effects for latent subgroups in observational studies. Stat Med. 2019;38(3):339–53.
Nielsen AM, Hestbaek L, Vach W, Kent P, Kongsted A. Latent class analysis derived subgroups of low back pain patients -do they have prognostic capacity? BMC Musculoskelet Disord. 2017;18(1):345. https://doi.org/10.1186/s12891-017-1708-9.
Nielsen AM, Kent P, Hestbaek L, Vach W, Kongsted A. Identifying subgroups of patients using latent class analysis: should we use a single-stage or a two-stage approach? a methodological study using a cohort of patients with low back pain. BMC Musculoskelet Disord. 2017;18(1):57. https://doi.org/10.1186/s12891-017-1411-x.
de Ruigh EL, Bouwmeester S, Popma A, Vermeiren RRJM, van Domburgh L, Jansen LMC. Using the biopsychosocial model for identifying subgroups of detained juveniles at different risk of re-offending in practice: a latent class regression analysis approach. Child Adolesc Psychiatry Ment Health. 2021;15(1):33. https://doi.org/10.1186/s13034-021-00379-1.
Shahn Z, Madigan D. Latent class mixture models of treatment effect heterogeneity. Bayesian Anal. 2017;12(3):831–54. https://doi.org/10.1214/16-BA1022.
Spilt JL, Koot JM, Lier PA. For whom does it work? subgroup differences in the effects of a school-based universal prevention program. Prev Sci. 2013;14:479–88. https://doi.org/10.1007/s11121-012-0329-7.
Willke RJ, Zheng Z, Subedi P, Althin R, Mullins CD. From concepts, theory, and evidence of heterogeneity of treatment effects to methodological approaches: a primer. BMC Med Res Methodol. 2012;12(1):185. https://doi.org/10.1186/1471-2288-12-185.
Woo JMP, Simanek A, O’Brien KM, Parks C, Gaston S, Auer PL, Konkel RH, Jackson CL, Meier HCS, Sandler DP. Latent class models of early-life trauma and incident breast cancer. Epidemiology. Accepted for publication. https://doi.org/10.1097/EDE.0000000000001507.
Zhang Z, Abarda A, Contractor AA, Wang J, Dayton CM. Exploring heterogeneity in clinical trials with latent class analysis. Ann Transl Med. 2018;6(7). https://doi.org/10.21037/atm.2018.01.24.
Connors AF, Speroff T, Dawson NV, Thomas C, Harrell FE, Wagner D, Desbiens N, Goldman L, Wu AW, Califf RM, et al. The effectiveness of right heart catheterization in the initial care of critically iii patients. J Am Med Assoc. 1996;276(11):889–97.
Bray BC, Lanza ST, Tan X. Eliminating bias in classify-analyze approaches for latent class analysis. Struct Equ Model. 2015;22(1):1–11. https://doi.org/10.1080/10705511.2014.935265.
Gardner J. Identification and estimation of average causal effects when treatment status is ignorable within unobserved strata. Econ Rev. 2020;39(10):1014–41. https://doi.org/10.1080/07474938.2020.1735748.
Haviland AM, Nagin DS. Causal inferences with group based trajectory models. Psychometrika. 2005;70(3):557–8.
Haviland A, Nagin DS, Rosenbaum PR, Tremblay RE. Combining group-based trajectory modeling and propensity score matching for causal inferences in nonexperimental longitudinal data. Dev Psychol. 2008;44(2):422.
Kent P, Jensen RK, Kongsted A. A comparison of three clustering methods for finding subgroups in mri, sms or clinical data: Spss twostep cluster analysis, latent gold and snob. BMC Med Res Methodol. 2014;14(1):113. https://doi.org/10.1186/1471-2288-14-113.
Lanza ST, Rhoades BL. Latent class analysis: an alternative perspective on subgroup analysis in prevention and treatment. Prev Sci. 2013;14(2):157–68. https://doi.org/10.1007/s11121-011-0201-1.
Lanza ST, Tan X, Bray BC. Latent class analysis with distal outcomes: A flexible model-based approach. Structural Equation Modeling: A Multidisciplinary Journal. 2013;20(1):1–26. https://doi.org/10.1080/10705511.2013.742377.
Suk Y, Kim J-S, Kang H. Hybridizing machine learning methods and finite mixture models for estimating heterogeneous treatment effects in latent classes. J Educ Behav Stat. 2021;46(3):323–47.
Anandkumar A, Ge R, Hsu D, Kakade SM, Telgarsky M. Tensor decompositions for learning latent variable models. J Mach Learn Res. 2014;15:2773–832.
Louizos C, Shalit U, Mooij JM, Sontag D, Zemel R, Welling M. Causal effect inference with deep latent-variable models. Red Hook: Curran Associates Inc.; 2017. p. 6446–56.
Kurz CF, Hatfield LA. Identifying and interpreting subgroups in health care utilization data with count mixture regression models. Stat Med. 2019;38(22):4423–35. https://doi.org/10.1002/sim.8307.
Bartolucci F, Grilli L, Pieroni L. Estimating dynamic causal effects with unobserved confounders: a latent class version of the inverse probability weighted estimator. Technical report. Germany: University Library of Munich; 2012.
Bartolucci F, Grilli L, Pieroni L. Inverse probability weighting to estimate causal effects of sequential treatments: A latent class extension to deal with unobserved confounding. 46th Scientific Meeting of the Italian Statistical Society. Rome; 2012. ISBN 978-88-6129-882-8. https://www.sis-statistica.it/index.php?p=3985.
Kim J-S, Steiner PM, Lim W-C. Mixture modeling methods for causal inference with multilevel data. In: Harring JR, Stapleton LM, Beretvas SN, editors. Advances in Multilevel Modeling for Educational Research: Addressing Practical Issues Found in Real-World Applications. Charlotte: Information Age Publishing, Inc.; 2016. p. 335–59.
Hoeppner BB, Goodwin MS, Velicer WF, Mooney ME, Hatsukami DK. Detecting longitudinal patterns of daily smoking following drastic cigarette reduction. Addictive Behaviors. 2008;33(5):623–39. https://doi.org/10.1016/j.addbeh.2007.11.005.
Koo W, Kim H. Bayesian nonparametric latent class model for longitudinal data. Stat Methods Med Res. 2020;29(11):3381–95. https://doi.org/10.1177/0962280220928384.
Lin H, McCulloch CE, Turnbull BW, Slate EH, Clark LC. A latent class mixed model for analysing biomarker trajectories with irregularly scheduled observations. Stat Med. 2000;19(10):1303–18.
McCulloch CE, Lin H, Slate EH, Turnbull BW. Discovering subpopulation structure with latent class mixed models. Stat Med. 2002;21(3):417–29. https://doi.org/10.1002/sim.1027.
Goodman LA. The analysis of systems of qualitative variables when some of the variables are unobservable. Part I Modified Latent Struct Approach. 1974;79(5):1179–259.
Goodman LA. On the assignment of individuals to latent classes. Sociol Methodol. 2007;37(1):1–22. https://doi.org/10.1111/j.1467-9531.2007.00184.x.
Hagenaars JA, McCutcheon AL. Applied Latent Class Analysis. Cambridge University Press; 2002. https://doi.org/10.1017/CBO9780511499531.
McCutcheon AL. A Latent Class Analysis of Tolerance for Nonconformity in the American Public. Public Opin Q. 1985;49(4):474–88. https://doi.org/10.1086/268945.
McCutcheon AL. Latent Class Analysis, vol 64. Sage. 1987. https://doi.org/10.4135/9781412984713.
Vermunt JK, Magidson J. In: Van der Ark LA, Croon MA, Sijtsma K, editors. Factor Analysis With Categorical Indicators: A Comparison Between Traditional and Latent Class Approaches. Mahwah: Lawrence Erlbaum Associates Publishers; 2005. p. 41–62.
Fraley C, Raftery AE. Model-based clustering, discriminant analysis, and density estimation. J Am Stat Assoc. 2002;97(458):611–31. https://doi.org/10.1198/016214502760047131.
McLachlan GJ, Peel D. Finite Mixture Models. New York: Wiley; 2004.
Schlattmann P. Medical Applications of Finite Mixture Models. Berlin, Heidelberg: Springer; 2009. https://doi.org/10.1007/978-3-540-68651-4.
Bakk Z, Tekle FB, Vermunt JK. Estimating the association between latent class membership and external variables using bias-adjusted three-step approaches. J Am Stat Assoc. 2013;43(1):272–311. https://doi.org/10.1177/0081175012470644.
Dias J, Vermunt JK. A bootstrap-based aggregate classifier for model-based clustering. Comput Stat. 2008;23(4):643–59. https://doi.org/10.1007/s00180-007-0103-7.
Bolck A, Croon M, Hagenaars J. Estimating latent structure models with categorical variables: One-step versus three-step estimators. Political Anal. 2004;12(1):3–27. https://doi.org/10.1093/pan/mph001.
Vermunt JK. Latent class modeling with covariates: Two improved three-step approaches. Political Anal. 2010;18(4):450–69. https://doi.org/10.1093/pan/mpq025.
Bakk Z, Kuha J. Relating latent class membership to external variables: An overview. Br J Math Stat Psychol. 2021;74(2):340–62. https://doi.org/10.1111/bmsp.12227.
Masyn KE. Measurement invariance and differential item functioning in latent class analysis with stepwise multiple indicator multiple cause modeling. Struct Equ Model Multidiscip J. 2017;24(2):180–97. https://doi.org/10.1080/10705511.2016.1254049.
Vermunt JK, Magidson J. How to perform three-step latent class analysis in the presence of measurement non-invariance or differential item functioning. Struct Equ Model Multidiscip J. 2021;28(3):356–64. https://doi.org/10.1080/10705511.2020.1818084.
Carvalho C, Feller A, Murray J, Woody S, Yeager D. Assessing treatment effect variation in observational studies: Results from a data challenge. Observational Stud. 2019;5(1):21–35.
Dorie V, Hill J, Shalit U, Scott M, Cervone D. Automated versus Do-It-Yourself Methods for Causal Inference: Lessons Learned from a Data Analysis Competition. Stat Sci. 2019;34(1):43–68. https://doi.org/10.1214/18-STS667.
Vansteelandt S. Statistical modelling in the age of data science. Observational Stud. 2021;7(1):217–28.
Dias JG, Vermunt JK. Bootstrap methods for measuring classification uncertainty in latent class analysis. In: Rizzi A, Vichi M, editors. Compstat 2006 - Proceedings in Computational Statistics. Heidelberg: Physica-Verlag HD; 2006. p. 31–41. https://doi.org/10.1007/978-3-7908-1709-6_3.
Bang H, Robins JM. Doubly robust estimation in missing data and causal inference models. Biometrics. 2005;61(4):962–73. https://doi.org/10.1111/j.1541-0420.2005.00377.x.
Glynn AN, Quinn KM. An introduction to the augmented inverse propensity weighted estimator. Political Anal. 2010;18(1):36–56. https://doi.org/10.1093/pan/mpp036.
Robins JM, Rotnitzky A, Zhao LP. Estimation of regression coefficients when some regressors are not always observed. J Am Stat Assoc. 1994;89(427):846–66.
R Core Team. R: A Language and Environment for Statistical Computing. Vienna: R Foundation for Statistical Computing; 2020. https://www.R-project.org/.
Nylund-Gibson K, Masyn KE. Covariates and mixture modeling: Results of a simulation study exploring the impact of misspecified effects on class enumeration. Struct Equ Model Multidiscip J. 2016;23(6):782–97. https://doi.org/10.1080/10705511.2016.1221313.
Nylund-Gibson K, Grimm RP, Masyn KE. Prediction from latent classes: A demonstration of different approaches to include distal outcomes in mixture models. Struct Equ Model Multidiscip J. 2019;26(6):967–85. https://doi.org/10.1080/10705511.2019.1590146.
Akaike H. A new look at the statistical model identification. In: Selected Papers of Hirotugu Akaike. Springer; 1974. p. 215-22. https://doi.org/10.1007/978-1-4612-1694-0_16.
Schwarz G. Estimating the dimension of a model. Ann Stat. 1978;6(2):461–4. https://doi.org/10.1214/aos/1176344136.
Weller BE, Bowen NK, Faubert SJ. Latent class analysis: A guide to best practice. J Black Psychol. 2020;46(4):287–311. https://doi.org/10.1177/0095798420930932.
Rubin DB. The design versus the analysis of observational studies for causal effects: parallels with the design of randomized trials. Stat Med. 2007;26(1):20–36. https://doi.org/10.1002/sim.2739.
Rosenbaum PR, Rubin DB. The central role of the propensity score in observational studies for causal effects. Biometrika. 1983;70(1):41–55. https://doi.org/10.1093/biomet/70.1.41.
Lunceford JK, Davidian M. Stratification and weighting via the propensity score in estimation of causal treatment effects: a comparative study. Stat Med. 2004;23(19):2937–60. https://doi.org/10.1002/sim.1903.
Rosenbaum PR. Model-based direct adjustment. J Am Stat Assoc. 1987;82(398):387–94.
Kang JD, Schafer JL. Demystifying double robustness: A comparison of alternative strategies for estimating a population mean from incomplete data. Stat Sci. 2007;22(4):523–39.
Austin PC. An introduction to propensity score methods for reducing the effects of confounding in observational studies. Multivariate Behavioral Research. 2011;46(3):399–24. https://doi.org/10.1080/00273171.2011.568786.
Imai K, Ratkovic M. Covariate balancing propensity score. J R Stat Soc Ser B (Stat Methodol). 2014;76(1):243–63. https://doi.org/10.1111/rssb.12027.
Pregibon D. Resistant fits for some commonly used logistic models with medical applications. Biometrics. 1982;485-98. DOIurlhttps://doi.org/10.2307/2530463.
Petersen ML, Porter KE, Gruber S, Wang Y, Van Der Laan MJ. Diagnosing and responding to violations in the positivity assumption. Stat Methods Med Res. 2012;21(1):31–54.
Stürmer T, Webster-Clark M, Lund JL, Wyss R, Ellis AR, Lunt M, Rothman KJ, Glynn RJ. Propensity Score Weighting and Trimming Strategies for Reducing Variance and Bias of Treatment Effect Estimates: A Simulation Study. Am J Epidemiol. 2021;190(8):1659–70. https://doi.org/10.1093/aje/kwab041.
Zou H, Hastie T. Regularization and variable selection via the elastic net. J R Stat Soc Ser B (Stat Methodol). 2005;67(2):301–20.
Hoerl AE, Kennard RW. Ridge regression: Biased estimation for nonorthogonal problems. Technometrics. 1970;12(1):55–67. https://doi.org/10.1080/00401706.1970.10488634.
Tibshirani R. Regression shrinkage and selection via the LASSO: a retrospective. J R Stat Soc Ser B (Stat Methodol). 2011;73(3):273–82. https://doi.org/10.1111/j.1467-9868.2011.00771.x.
Friedman J, Hastie T, Tibshirani R. Regularization paths for generalized linear models via coordinate descent. J Stat Softw. 2010;33(1):1–22.
Westreich D, Cole SR. Invited Commentary: Positivity in Practice. American Journal of Epidemiology. 2010;171(6):674–7. https://doi.org/10.1093/aje/kwp436.
Bandeen-Roche K, Miglioretti DL, Zeger SL, Rathouz PJ. Latent variable regression for multiple discrete outcomes. J Am Stat Assoc. 1997;92(440):1375–86.
Vansteelandt S, Goetghebeur E, Kenward MG, Molenberghs G. Ignorance and uncertainty regions as inferential tools in a sensitivity analysis. Stat Sin. 2006;953-79.
Liu Y, Yang JS. Interval estimation of latent variable scores in item response theory. Journal of Educational and Behavioral Statistics. 2018;43(3):259–85.
Yang JS, Hansen M, Cai L. Characterizing sources of uncertainty in item response theory scale scores. Educ Psychol Meas. 2012;72(2):264–90.
Bakk Z, Oberski DL, Vermunt JK. Relating latent class membership to continuous distal outcomes: Improving the ltb approach and a modified three-step implementation. Struct Equ Model Multidiscip J. 2016;23(2):278–89. https://doi.org/10.1080/10705511.2015.1049698.
Helmreich JE, Pruzek RM. PSAgraphics: An R package to support propensity score analysis. J Stat Softw. 2009;29(6):1–23.
Cefalu M, Ridgeway G, McCaffrey D, Morral A, Griffin BA, Burgette L. Twang: Toolkit for Weighting and Analysis of Nonequivalent Groups. R package version 2.0. 2021. https://CRAN.R-project.org/package=twang.
Linzer D.A, Lewis JB, et al. poLCA: An R package for polytomous variable latent class analysis. Journal of statistical software. 2011;42(10):1-29. https://doi.org/10.18637/jss.v042.i10.
Dayton CM, Macready GB. Concomitant-variable latent-class models. J Am Stat Assoc. 1988;83(401):173–8. https://doi.org/10.1080/01621459.1988.10478584.
Bollen K, Lennox R. Conventional wisdom on measurement: A structural equation perspective. Psychol Bull. 1991;110(2):305.
Vermunt JK, Magidson J. Technical guide for latent gold 5.0: Basic, advanced, and syntax. Statistical Innovations Inc: Belmont; 2013.
Vermunt J, Magidson J. Upgrade manual for Latent GOLD 6.0. Statistical Innovations Inc; 2020.
Lanza ST, Coffman DL, Xu S. Causal inference in latent class analysis. Psychol Bull Multidiscip J. 2013;20(3):361–83. https://doi.org/10.1080/10705511.2013.797816.
Clouth FJ, Pauws S, Mols F, Vermunt JK. A new three-step method for using inverse propensity weighting with latent class analysis. Adv Data Anal Classif. 2021. https://doi.org/10.1007/s11634-021-00456-5.
Bray BC, Dziak JJ, Patrick ME, Lanza ST. Inverse propensity score weighting with a latent class exposure: Estimating the causal effect of reported reasons for alcohol use on problem alcohol use 16 years later. Prev Sci. 2019;20(3):394–406. https://doi.org/10.1007/s11121-018-0883-8.
Schuler MS, Leoutsakos JS, Stuart EA. Addressing confounding when estimating the effects of latent classes on a distal outcome. Health Serv Outcomes Res Methodol. 2014;14:232–54. https://doi.org/10.1007/s10742-014-0122-0.
Bakk Z, Vermunt JK. Robustness of stepwise latent class modeling with continuous distal outcomes. Struct Equ Model. 2016;23(1):20–31. https://doi.org/10.1080/10705511.2014.955104.
Asparouhov T, Muthén B. Auxiliary variables in mixture modeling: Three-step approaches using Mplus. Struct Equ Model. 2014;21(3):329–41. https://doi.org/10.1080/10705511.2014.915181.
Mayer A, Zimmermann J, Hoyer J, Salzer S, Wiltink J, Leibing E, Leichsenring F. Interindividual differences in treatment effects based on structural equation models with latent variables: An EffectLiteR tutorial. Struct Equ Model Multidiscip J. 2019;1-19. https://doi.org/10.1080/10705511.2019.1671196.
Jacob D. Cross-fitting and averaging for machine learning estimation of heterogeneous treatment effects. arXiv preprint arXiv:2007.02852. 2020.
Grün B, Leisch F. FlexMix version 2: Finite mixtures with concomitant variables and varying and constant parameters. J Stat Softw. 2008;28(4):1-35. https://doi.org/10.18637/jss.v028.i04.
Jakobsen JC, Gluud C, Wetterslev J, Winkel P. When and how should multiple imputation be used for handling missing data in randomised clinical trials - a practical guide with flowcharts. BMC Med Res Methodol. 2017;17(1):162. https://doi.org/10.1186/s12874-017-0442-1.
Bartolucci F, Bacci S, Gnaldi M. MultiLCIRT: An R package for multidimensional latent class item response models. Comput Stat Data Anal. 2014;71:971–85. https://doi.org/10.1016/j.csda.2013.05.018.
Gemma M, Pennoni F, Braga M. Studying enhanced recovery after surgery (eras®) core items in colorectal surgery: A causal model with latent variables. World J Surg. 2021;45(4):928–39. https://doi.org/10.1007/s00268-020-05940-1.
McLachlan GJ, Lee SX, Rathnayake SI. Finite mixture models. Ann Rev Stat Appl. 2019;6(1):355–78. https://doi.org/10.1146/annurev-statistics-031017-100325.
Teicher H, et al. On the mixture of distributions. Ann Math Stat. 1960;31(1):55–73. https://doi.org/10.1214/aoms/1177705987.
Austin PC. Assessing the performance of the generalized propensity score for estimating the effect of quantitative or continuous exposures on binary outcomes. Stat Med. 2018;37(11):1874–94. https://doi.org/10.1002/sim.7615.
Bartolucci F, Pennoni F, Vittadini G. Causal latent markov model for the comparison of multiple treatments in observational longitudinal studies. J Educ Behav Stat. 2016;41(2):146–79. https://doi.org/10.3102/1076998615622234.
Joffe MM, Yang WP, Feldman HI. Selective ignorability assumptions in causal inference. Int J Biostat. 2010;6(2).
Pal N.R, Pal K, Keller JM, Bezdek JC. A possibilistic fuzzy c-means clustering algorithm. IEEE Trans Fuzzy Syst. 2005;13(4):517-30. https://doi.org/10.1109/TFUZZ.2004.840099.