{"title":"dynr.mi: An R Program for Multiple Imputation in Dynamic Modeling","authors":"Yanling Li, Linying Ji, Zita Oravecz, Timothy R. Brick, Michael D. Hunter, Sy-Miin Chow","volume":149,"journal":"International Journal of Computer and Information Engineering","pagesStart":298,"pagesEnd":308,"ISSN":"1307-6892","URL":"https:\/\/publications.waset.org\/pdf\/10010406","abstract":"Assessing several individuals intensively over time
\r\nyields intensive longitudinal data (ILD). Even though ILD provide
\r\nrich information, they also bring other data analytic challenges. One
\r\nof these is the increased occurrence of missingness with increased
\r\nstudy length, possibly under non-ignorable missingness scenarios.
\r\nMultiple imputation (MI) handles missing data by creating several
\r\nimputed data sets, and pooling the estimation results across imputed
\r\ndata sets to yield final estimates for inferential purposes. In this
\r\narticle, we introduce dynr.mi(), a function in the R package,
\r\nDynamic Modeling in R (dynr). The package dynr provides a suite
\r\nof fast and accessible functions for estimating and visualizing the
\r\nresults from fitting linear and nonlinear dynamic systems models in
\r\ndiscrete as well as continuous time. By integrating the estimation
\r\nfunctions in dynr and the MI procedures available from the R
\r\npackage, Multivariate Imputation by Chained Equations (MICE), the
\r\ndynr.mi() routine is designed to handle possibly non-ignorable
\r\nmissingness in the dependent variables and\/or covariates in a
\r\nuser-specified dynamic systems model via MI, with convergence
\r\ndiagnostic check. We utilized dynr.mi() to examine, in the context
\r\nof a vector autoregressive model, the relationships among individuals’
\r\nambulatory physiological measures, and self-report affect valence
\r\nand arousal. The results from MI were compared to those from
\r\nlistwise deletion of entries with missingness in the covariates.
\r\nWhen we determined the number of iterations based on the
\r\nconvergence diagnostics available from dynr.mi(), differences in
\r\nthe statistical significance of the covariate parameters were observed
\r\nbetween the listwise deletion and MI approaches. These results
\r\nunderscore the importance of considering diagnostic information in
\r\nthe implementation of MI procedures.","references":"[1] N. Bolger and J.-P. Laurenceau, Intensive Longitudinal Methods: An\r\nIntroduction to Diary and Experience Sampling Research. New York,\r\nNY: Guilford Press, 2013.\r\n[2] T. H. Walls and J. L. Schafer, Models for intensive longitudinal data.\r\nOxford: University Press, 2006.\r\n[3] S. Shiffman, A. A. Stone, and M. R. Hufford, \u201cEcological momentary\r\nassessment.\u201d Annual Review of Clinical Psychology, vol. 4, pp. 1\u201332,\r\n2008.\r\n[4] A. Stone, S. Shiffman, A. Atienza, and L. Nebeling, The Science of\r\nReal-Time Data Capture: Self-Reports in Health Research. NY: Oxford\r\nUniversity Press, 2008.\r\n[5] L. Ou, M. D. Hunter, and S.-M. Chow, \u201cWhat\u2019s for dynr: A package\r\nfor linear and nonlinear dynamic modeling in R,\u201d The R Journal, 2019,\r\nin press.\r\n[6] D. B. Rubin, \u201cInference and missing data,\u201d Biometrika, vol. 63, no. 3,\r\npp. 581\u2013592, 1976.\r\n[7] L. Ji, S.-M. Chow, A. C. Schermerhorn, N. C. Jacobson, and E. M.\r\nCummings, \u201cHandling missing data in the modeling of intensive\r\nlongitudinal data,\u201d Structural Equation Modeling: A Multidisciplinary\r\nJournal, pp. 1\u201322, 2018.\r\n[8] D. B. Rubin, Multiple imputation for nonresponse in surveys. John\r\nWiley & Sons, 2004, vol. 81.\r\n[9] S. van Buuren and C. Oudshoorn, \u201cMultivariate imputation by chained\r\nequations,\u201d MICE V1. 0 user\u2019s manual. Leiden: TNO Preventie en\r\nGezondheid, 2000.\r\n[10] S. van Buuren and K. Groothuis-Oudshoorn, \u201cmice: Multivariate\r\nimputation by chained equations in R,\u201d Journal of Statistical\r\nSoftware, vol. 45, no. 3, pp. 1\u201367, 2011. [Online]. Available:\r\nhttp:\/\/www.jstatsoft.org\/v45\/i03\/\r\n[11] T. E. Raghunathan, J. M. Lepkowski, J. Van Hoewyk, P. Solenberger\r\net al., \u201cA multivariate technique for multiply imputing missing values\r\nusing a sequence of regression models,\u201d Survey methodology, vol. 27,\r\nno. 1, pp. 85\u201396, 2001.\r\n[12] T. W. Anderson, \u201cMaximum likelihood estimates for a multivariate\r\nnormal distribution when some observations are missing,\u201d Journal of\r\nthe American Statistical Association, vol. 52, pp. 200\u2013203, June 1957.\r\n[Online]. Available: http:\/\/www.jstor.org\/stable\/2280845\r\n[13] J. A. Russell, \u201cCore affect and the psychological construction of\r\nemotion,\u201d Psychological Review, vol. 110, pp. 145\u2013172, 2003.\r\n[14] P. Kuppens, Z. Oravecz, and F. Tuerlinckx, \u201cFeelings change:\r\nAccounting for individual differences in the temporal dynamics of\r\naffect,\u201d Journal of Personality and Social Psychology, vol. 99, pp.\r\n1042\u20131060, 2010.\r\n[15] U. Ebner-Priemer, M. Houben, P. Santangelo, N. Kleindienst,\r\nF. Tuerlinckx, Z. Oravecz, G. Verleysen, K. V. Deun, M. Bohus, and\r\nP. Kuppens, \u201cUnraveling affective dysregulation in borderline personality\r\ndisorder: a theoretical model and empirical evidence.\u201d Journal of\r\nabnormal psychology, vol. 124 1, pp. 186\u201398, 2015.\r\n[16] J. A. Russell and L. F. Barrett, \u201cCore affect, prototypical emotional\r\nepisodes, and other things called emotion: dissecting the elephant,\u201d\r\nJournal of Personality and Social Psychology, vol. 76, pp. 805\u2013819,\r\n1999.\r\n[17] R. W. Picard, S. Fedor, and Y. Ayzenberg, \u201cMultiple Arousal\r\nTheory and Daily-Life Electrodermal Activity Asymmetry,\u201d\r\nEmotion Review, pp. 1\u201314, Mar. 2015. [Online]. Available:\r\nhttp:\/\/emr.sagepub.com\/content\/early\/2015\/02\/20\/1754073914565517\r\n[18] N. L. Sin, R. P. Sloan, P. S. McKinley, and D. M. Almeida, \u201cLinking\r\ndaily stress processes and laboratory-based heart rate variability in a\r\nnational sample of midlife and older adults,\u201d Psychosomatic medicine,\r\nvol. 78(5), pp. 573\u2013582, 2016.\r\n[19] T. Bossmann, M. K. Kanning, S. Koudela-Hamila, S. Hey, and\r\nU. Ebner-Priemer, \u201cThe association between short periods of everyday\r\nlife activities and affective states: A replication study using ambulatory\r\nassessment,\u201d Frontiers in Psychology, vol. 4, 2013.\r\n[20] G. F. Dunton, J. Huh, A. M. Leventhal, N. R. Riggs, D. Hedeker,\r\nD. Spruijt-Metz, and M. A. Pentz, \u201cMomentary assessment of affect,\r\nphysical feeling states, and physical activity in children.\u201d Health\r\npsychology : official journal of the Division of Health Psychology,\r\nAmerican Psychological Association, vol. 33 3, pp. 255\u201363, 2014. [21] M. K. Kanning and D. Schoebi, \u201cMomentary affective states are\r\nassociated with momentary volume, prospective trends, and fluctuation\r\nof daily physical activity,\u201d Frontiers in Psychology, vol. 7, 2016.\r\n[22] C. Y. N. Niermann, C. Herrmann, B. von Haaren, D. H. H. V. Kann,\r\nand A. Woll, \u201cAffect and subsequent physical activity: An ambulatory\r\nassessment study examining the affect-activity association in a real-life\r\ncontext,\u201d Frontiers in Psychology, vol. 7, 2016.\r\n[23] Y. Liao, C.-P. Chou, J. Huh, A. M. Leventhal, and G. F. Dunton,\r\n\u201cExamining acute bi-directional relationships between affect, physical\r\nfeeling states, and physical activity in free-living situations using\r\nelectronic ecological momentary assessment,\u201d Journal of Behavioral\r\nMedicine, vol. 40, pp. 445\u2013457, 2016.\r\n[24] S.-M. Chow, M.-H. R. Ho, E. J. Hamaker, and C. V. Dolan,\r\n\u201cEquivalences and differences between structural equation and\r\nstate-space modeling frameworks,\u201d Structural Equation Modeling,\r\nvol. 17, pp. 303\u2013332, 2010.\r\n[25] J. Durbin and S. J. Koopman, Time Series Analysis by State Space\r\nMethods. Oxford, United Kingdom: Oxford University Press, 2001.\r\n[26] S.-M. Chow, L. Ou, A. Ciptadi, E. Prince, M. D. Hunter, D. You, J. M.\r\nRehg, A. Rozga, and D. S. Messinger, \u201cRepresenting sudden shifts in\r\nintensive dyadic interaction data using differential equation models with\r\nregime switching,\u201d Psychometrika, vol. 83, no. 2, pp. 476\u2013510, 2018.\r\n[27] R. E. Kalman, \u201cA new approach to linear filtering and prediction\r\nproblems,\u201d Journal of Basic Engineering, vol. 82, no. 1, pp. 35\u201345,\r\n1960.\r\n[28] P. De Jong, \u201cThe likelihood for a state space model,\u201d Biometrika, vol. 75,\r\nno. 1, pp. 165\u2013169, March 1988.\r\n[29] A. C. Harvey, Forecasting, Structural Time Series Models and the\r\nKalman Filter. Cambridge, United Kingdom: Cambridge University\r\nPress, 1989.\r\n[30] J. D. Hamilton, Time Series Analysis. Princeton, NJ: Princeton\r\nUniversity Press, 1994.\r\n[31] S.-M. Chow and G. Zhang, \u201cNonlinear regime-switching state-space\r\n(RSSS) models,\u201d Psychometrika, vol. 78, no. 4, pp. 740\u2013768, 2013.\r\n[32] H. Akaike, \u201cInformation theory and an extension of the maximum\r\nlikelihood principle,\u201d in Second International Symposium on Information\r\nTheory, B. N. Petrov and F. Csaki, Eds. Budapest: Akademiai Kiado,\r\n1973, pp. 267\u2013281.\r\n[33] G. Schwarz, \u201cEstimating the dimension of a model,\u201d The Annals of\r\nStatistics, vol. 6, no. 2, pp. 461\u2013464, 1978.\r\n[34] F. Thoemmes and N. Rose, \u201cA cautious note on auxiliary variables that\r\ncan increase bias in missing data problems,\u201d Multivariate Behavioral\r\nResearch, vol. 49, no. 5, pp. 443\u2013459, 2014.\r\n[35] L. M. Collins, J. L. Schafer, and C.-M. Kam, \u201cA comparison of\r\ninclusive and restrictive strategies in modern missing data procedures.\u201d\r\nPsychological methods, vol. 6, no. 4, p. 330, 2001.\r\n[36] A. Gelman and D. B. Rubin, \u201cInference from iterative simulation using\r\nmultiple sequences,\u201d Statistical science, vol. 7, no. 4, pp. 457\u2013472, 1992.\r\n[37] S. Brooks and A. Gelman, \u201cSome issues for monitoring convergence\r\nof iterative simulations,\u201d Computing Science and Statistics, pp. 30\u201336,\r\n1998.\r\n[38] R. W. Picard, \u201cRecognizing Stress, Engagement, and Positive Emotion,\u201d\r\nin Proceedings of the 20th International Conference on Intelligent User\r\nInterfaces, ser. IUI \u201915. New York, NY, USA: ACM, 2015, pp. 3\u20134.\r\n[Online]. Available: http:\/\/doi.acm.org\/10.1145\/2678025.2700999\r\n[39] P. Kuppens, N. B. Allen, and L. B. Sheeber, \u201cEmotional inertia and\r\npsychological maladjustment,\u201d Psychological Science, 2010.\r\n[40] P. Royston, \u201cMultiple imputation of missing values,\u201d The Stata Journal,\r\nvol. 4, no. 3, pp. 227\u2013241, 2004.\r\n[41] K. Lu, \u201cNumber of imputations needed to stabilize estimated treatment\r\ndifference in longitudinal data analysis,\u201d Statistical methods in medical\r\nresearch, vol. 26, no. 2, pp. 674\u2013690, 2017.\r\n[42] R. J. Little and D. B. Rubin, Statistical analysis with missing data.\r\nWiley, 2019, vol. 793.","publisher":"World Academy of Science, Engineering and Technology","index":"Open Science Index 149, 2019"}