Inversion Theorem Based Kernel Density Estimation for the Ordinary Least Squares Estimator of a Regression Coefficient

The traditional confidence interval associated with the ordinary least squares estimator of linear regression coefficient is sensitive to non-normality of the underlying distribution. In this article, we develop a novel kernel density estimator for the ordinary least squares estimator via utilizing well-defined inversion based kernel smoothing techniques in order to estimate the conditional probability density distribution of the dependent random variable. Simulation results show that given a small sample size, our method significantly increases the power as compared with Wald-type CIs. The proposed approach is illustrated via an application to a classic small data set originally from Graybill (1961 Graybill, F.A. (1961). Introduction to Linear Statistical Models. Vol. 1. New York: McGraw-Hill Book Company. [Google Scholar]).     

Some Classes of Estimators for the Population Median Using Auxiliary Information

This article considers some classes of estimators of the population median of the study variable using information on an auxiliary variable with their properties under large sample approximation. Asymptotic optimum estimator (AOE) in each class of estimators has been investigated along with the approximate mean square error formulae. It has been shown that the proposed classes of estimators are better than these considered by Gross (1980 Gross , T. S. ( 1980 ). Median estimation in sample surveys. Proc. Surv. Res. Meth. Sect. Amer. Statist. Assoc. 181–184 . [Google Scholar]), Kuk and Mak (1989 Kuk , A. Y. C. , Mak , T. K. ( 1989 ). Median estimation in the presence of auxiliary information . J. Roy. Statist. Soc. Ser. B51 : 261 – 269 . [Google Scholar]), Singh et al. (2003a Singh , H. P. , Singh , S. , Joarder , A. H. ( 2003a ). Estimation of population median when mode of an auxiliary variable is known . J. Statist. Res. 37 ( 1 ): 57 – 63 . [Google Scholar]), and Al and Cingi (2009 Al , S. , Cingi , H. ( 2009 ). New estimators for the population median in simple random sampling. Tenth Islamic Countries Conference on Statistical Sciences, held in New Cairo, Egypt . [Google Scholar]). An empirical study is carried out to judge the merits of the suggested class of estimators over other existing estimators.     

Some new methods to solve multicollinearity in logistic regression

The binary logistic regression is a widely used statistical method when the dependent variable is binary or dichotomous. In some of the situations of logistic regression, independent variables are collinear which leads to the problem of multicollinearity. It is known that multicollinearity affects the variance of maximum likelihood estimator (MLE) negatively. Thus, this article introduces new methods to estimate the shrinkage parameters of Liu-type logistic estimator proposed by Inan and Erdogan (2013 Inan, D., Erdogan, B. E. (2013). Liu-type logistic estimator. Communications in Statistics-Simulation and Computation 42(7):1578–1586. [Google Scholar]) which is a generalization of the Liu-type estimator defined by Liu (2003 Liu, K. (2003). Using Liu-type estimator to combat collinearity. Communications in Statistics: Theory and Methods 32(5):1009–1020. [Google Scholar]) for the linear model. A Monte Carlo study is used to show the effectiveness of the proposed methods over MLE using the mean squared error (MSE) and mean absolute error (MAE) criteria. A real data application is illustrated to show the benefits of new methods. According to the results of the simulation and application proposed methods have better performance than MLE.     

Relative Errors of Difference-Based Variance Estimators in Nonparametric Regression

Difference-based estimators for the error variance are popular since they do not require the estimation of the mean function. Unlike most existing difference-based estimators, new estimators proposed by Müller et al. (2003 Müller , U. , Schick , A. , Wefelmeyer , W. ( 2003 ). Estimating the error variance in nonparametric regression by a covariate-matched U-statistic . Statistics 37 : 179 – 188 .[Taylor & Francis Online], [Web of Science ®] , [Google Scholar]) and Tong and Wang (2005 Tong , T. , Wang , Y. ( 2005 ). Estimating residual variance in nonparametric regression using least squares . Biometrika 92 : 821 – 830 .[Crossref], [Web of Science ®] , [Google Scholar]) achieved the asymptotic optimal rate as residual-based estimators. In this article, we study the relative errors of these difference-based estimators which lead to better understanding of the differences between them and residual-based estimators. To compute the relative error of the covariate-matched U-statistic estimator proposed by Müller et al. (2003 Müller , U. , Schick , A. , Wefelmeyer , W. ( 2003 ). Estimating the error variance in nonparametric regression by a covariate-matched U-statistic . Statistics 37 : 179 – 188 .[Taylor & Francis Online], [Web of Science ®] , [Google Scholar]), we develop a modified version by using simpler weights. We further investigate its asymptotic property for both equidistant and random designs and show that our modified estimator is asymptotically efficient.     

Goodness-of-Fit Test and Parameter Estimation for a Proportional Odds Model of Random Censorship

Censored data arise naturally in a number of fields, particularly in problems of reliability and survival analysis. There are several types of censoring, in this article, we will confine ourselves to the right randomly censoring type. Recently, Ahmadi et al. (2010 Ahmadi , J. , Doostparast , M. , Parsian , A. ( 2010 ). Bayes estimation based on random censored data for some life time models under symmetric and asymmetric loss functions . Communcations in Statistics-Theory and Methods , 39 : 3058 – 3071 .[Taylor & Francis Online], [Web of Science ®] , [Google Scholar]) considered the problem of estimating unknown parameters in a general framework based on the right randomly censored data. They assumed that the survival function of the censoring time is free of the unknown parameter. This assumption is sometimes inappropriate. In such cases, a proportional odds (PO) model may be more appropriate (Lam and Leung, 2001 Lam , K. F. , Leung , T. L. ( 2001 ). Marginal likelihood estimation for proportional odds models with right censored data . Lifetime Data Analysis 7 : 39 – 54 .[Crossref], [PubMed], [Web of Science ®] , [Google Scholar]). Under this model, in this article, point and interval estimations for the unknown parameters are obtained. Since it is important to check the adequacy of models upon which inferences are based (Lawless, 2003 Lawless , J. F. (2003). Statistical Models and Methods for Lifetime Data. , 2nd ed. New York : John Wiley & Sons. [Google Scholar], p. 465), two new goodness-of-fit tests for PO model based on right randomly censored data are proposed. The proposed procedures are applied to two real data sets due to Smith (2002 Smith , P. J. ( 2002 ). Analysis of Failure and Survival Data . London : Chapman & Hall, CRC . [Google Scholar]). A Monte Carlo simulation study is conducted to carry out the behavior of the estimators obtained.     

ESTIMATING PARAMETERS IN AUTOREGRESSIVE MODELS IN NON-NORMAL SITUATIONS: ASYMMETRIC INNOVATIONS

The estimation of coefficients in a simple autoregressive model is considered in a supposedly difficult situation where the innovations have an asymmetric distribution. Two distributions, gamma and generalized logistic, are considered for illustration. Closed form estimators are obtained and shown to be efficient and robust. Efficiencies of least squares estimators are evaluated and shown to be very low. This work is an extension of that of Tiku, Wong and Bian [1] Tiku, M. L., Wong, W. K. and Bian, G. 1999. Time Series Models with Asymmetric Innovations. Commun. Stat.-Theory Meth., 28: 1131–1160.  [Google Scholar] who give solutions for a simple AR(1) model.

     

A Consistent Variance Estimator for 2SLS When Instruments Identify Different LATEs

Under treatment effect heterogeneity, an instrument identifies the instrument-specific local average treatment effect (LATE). With multiple instruments, two-stage least squares (2SLS) estimand is a weighted average of different LATEs. What is often overlooked in the literature is that the postulated moment condition evaluated at the 2SLS estimand does not hold unless those LATEs are the same. If so, the conventional heteroscedasticity-robust variance estimator would be inconsistent, and 2SLS standard errors based on such estimators would be incorrect. I derive the correct asymptotic distribution, and propose a consistent asymptotic variance estimator by using the result of Hall and Inoue (2003 Hall, A.R., and Inoue, A. (2003), “The Large Sample Behaviour of the Generalized Method of Moments Estimator in Misspecified Models,” Journal of Econometrics, 114, 361–394.[Crossref], [Web of Science ®] , [Google Scholar], Journal of Econometrics) on misspecified moment condition models. This can be used to correctly calculate the standard errors regardless of whether there is more than one LATE or not.     

Statistical Properties of Threshold Models

This article focuses the attention on the Self Exciting Threshold Autoregressive Moving Average model (SETARMA) proposed in Tong (1983 Tong , H. ( 1983 ). Threshold Models in Nonlinear Time Series Analysis . London : Springer-Verlag .[Crossref] , [Google Scholar]). The stochastic structure of the model is discussed and different specifications are presented. Starting from one of them, we give sufficient conditions for the weak stationarity of the model that are discussed and critically compared to other results given in literature. In particular, after showing that the SETARMA model belongs to the class of the Random Coefficients Autoregressive models, widely discussed in Nicholls and Quinn (1982 Nicholls , D. F. , Quinn , B. G. (1982). Random Coefficients Autoregressive Models. An Introduction . New York : Springer-Verlag.[Crossref] , [Google Scholar]), we give some issues on the weak stationarity of its stochastic structure that are more general than those given in the existing literature and appear not affected by the moving average component.     

SECOND-ORDER ASYMPTOTICS FOR SCORE TESTS IN HETEROSKEDASTIC t REGRESSION MODELS

ABSTRACT

This paper develops corrected score tests for heteroskedastic t regression models, thus generalizing results by Cordeiro, Ferrari and Paula[1] Cordeiro, G.M., Ferrari, S.L.P. and Paula, G.A. 1993. Improved Score Tests for Generalized Linear Models. Journal of the Royal Statistical Society B, 55: 661–674.  [Google Scholar] and Cribari-Neto and Ferrari[2] Cribari-Neto, F. and Ferrari, S.L.P. 1995. Second-order Asymptotics for Score Tests in Generalised Linear Models. Biometrika, 82: 426–432. [Crossref], [Web of Science ®] , [Google Scholar] for normal regression models and by Ferrari and Arellano-Valle[3] Ferrari, S.L.P. and Arellano-Valle, R. 1996. Modified Likelihood Ratio and Score Tests in Linear Regression Models Using the t Distribution. Brazilian Journal of Probability and Statistics, 10: 15–33.  [Google Scholar] for homoskedastic t regression models. We present, in matrix notation, Bartlett-type correction formulae to improve score tests in this class of models. The corrected score statistics have a chi-squared distribution to order n ^?1, where n is the sample size. We apply our main result to a few special models and present simulation results comparing the performance of the usual score tests and their corrected versions.     

Estimation of Parameters of Mixed Failure Time Distribution Based on an Extended Modified Sampling Scheme

Abstract

The problem of estimation of parameters of a mixture of degenerate (at zero) and exponential distribution is considered by Dixit and Prasad [Dixit, V. U. (Nee: Jayade, V. D.), Prasad, M. S. (1990 Dixit, V. U. and Prasad, M. S. 1990. Estimation of parameters of mixed failure time distribution. Commun.in Statist.-Theory Meth, 19(12): 4667–4678. (Nee: Jayade, V. D.) [Google Scholar]). Estimation of parameters of mixed failure time distribution. Commun.in Statist.-Theory Meth., 19(12):4667–4678]. The sampling scheme proposed in it is extended to k positive observations in Dixit [Dixit, V. U. (1993 Dixit, V. U. 1993. “Statistical Inference for AR (1) Process with Mixed Errors”. Kolhapur, , India: Shivaji University. Unpublished Ph.D. thesis [Google Scholar]). Statistical Inference for AR (1) Process with Mixed Errors. Unpublished Ph.D. thesis, Shivaji University Kolhapur, India] and moment estimator, MLE and UMVUE based on it are obtained and their finite sample and asymptotic properties are studied. These results are presented in this paper. It is interesting to mention that the sampling scheme proposed by Shinde and Shanubhogue [Shinde, R. L., Shanubhogue, A. (2000 Shinde, R. L. and Shanubhogue, A. 2000. Estimation of parameters and the mean life of a mixed failure time distribution. Commun. Statist.-Theory Meth, 29(11): 2621–2642. [Taylor & Francis Online], [Web of Science ®] , [Google Scholar]). Estimation of parameters and the mean life of a mixed failure time distribution. Commun. Statist.-Theory Meth. 29(11):2621–2642] is a particular case of the sampling scheme proposed in Dixit [Dixit, V. U. (1993 Dixit, V. U. 1993. “Statistical Inference for AR (1) Process with Mixed Errors”. Kolhapur, , India: Shivaji University. Unpublished Ph.D. thesis [Google Scholar]). Statistical Inference for AR (1) Process with Mixed Errors. Unpublished Ph.D. thesis, Shivaji University Kolhapur, India] for n = k.     

On the optimal allocation of active redundancies in series system

This paper studies the allocations of two non identical active redundancies in series systems in terms of the reversed hazard rate order and hazard rate order, which generalizes some results built in Valdés and Zequeira (2003 Valdés, J. E., and R. I. Zequeira 2003. On the optimal allocation of an active redundancy in a two-component series system. Stat. Probab. Lett. 63:325–32.[Crossref], [Web of Science ®] , [Google Scholar], 2006 Valdés, J. E., and R. I. Zequeira 2006. On the optimal allocation of two active redundancies in a two-component series system. Oper. Res. Lett. 34:49–52.[Crossref], [Web of Science ®] , [Google Scholar]).     

Standard Exponential Cure Rate Model with Informative Censoring

In this article, we consider the standard cure rate model proposed by Boag (1949 Boag , J. W. ( 1949 ). Maximum likelihood estimates of the proportion of patients cured by cancer therapy . J. Roy. Statist. Soc. B 11 : 15 – 53 . [Google Scholar]) and Berkson and Gage (1952 Berkson , J. , Gage , R. ( 1952 ). Survival curve for cancer patients following treatment . J. Amer. Statist. Assoc. 47 : 501 – 515 .[Taylor & Francis Online], [Web of Science ®] , [Google Scholar]). We present a new definition of informative censoring similar to Lawless (1982 Lawless , J. F. ( 1982 ). Statistical Models and Methods for Lifetime Data . New York : Wiley .[Crossref] , [Google Scholar]) and the corresponding likelihood function. Under informative censoring, we obtain the Fisher information matrix of the exponential standard cure rate model. We verify, with simulated data, the impact caused by informative censoring in the coverage probabilities and in the lengths of asymptotic confidence intervals of the parameters of interest. An example with real data is analyzed.     

A quantile-based study on ageing intensity function

Jiang, Ji, and Xiao (2003 Jiang, R., P. Ji, and X. Xiao. 2003. Aging property of unimodal failure rate models. Reliability Engineering and System Safety 79(1):113–6.[Crossref], [Web of Science ®] , [Google Scholar]) has introduced a quantitative measure known as the ageing intensity function for evaluating the ageing properties of a component/system. In recent years, there has been a great interest on the study of quantile function, an equivalent alternative to the distribution function approach. Unlike the distribution function approach, the quantile method possess some unique properties (see Gilchrist 2000 Gilchrist, W. 2000. Statistical modelling with quantile functions. Boca Raton, Florida: Chapman and Hall/CRC.[Crossref] , [Google Scholar], Nair, Sankaran, and Balakrishnan 2013 Nair N. U., P. G. Sankaran, N. Balakrishnan. 2013. Quantile-based reliability concepts. In: Quantile-Based Reliability Analysis. Statistics for Industry and Technology. New York, NY: Birkhäuser.[Crossref] , [Google Scholar]). Motivated with this, in the present paper we introduce a quantile-based ageing intensity function and study its various ageing properties. We also study some stochastic comparison of random variables based on the proposed measure.     

Measures of predictor sensitivity for order-insensitive partitioning of multiple correlation

Lindeman et al. [12 Lindeman, R. H., Merenda, P. F. and Gold, R. Z. 1980. Introduction to Bivariate and Multivariate Analysis, Glenview, IL: Scott Foresman.  [Google Scholar]] provide a unique solution to the relative importance of correlated predictors in multiple regression by averaging squared semi-partial correlations obtained for each predictor across all p! orderings. In this paper, we propose a series of predictor sensitivity statistics that complement the variance decomposition procedure advanced by Lindeman et al. [12 Lindeman, R. H., Merenda, P. F. and Gold, R. Z. 1980. Introduction to Bivariate and Multivariate Analysis, Glenview, IL: Scott Foresman.  [Google Scholar]]. First, we detail the logic of averaging over orderings as a technique of variance partitioning. Second, we assess predictors by conditional dominance analysis, a qualitative procedure designed to overcome defects in the Lindeman et al. [12 Lindeman, R. H., Merenda, P. F. and Gold, R. Z. 1980. Introduction to Bivariate and Multivariate Analysis, Glenview, IL: Scott Foresman.  [Google Scholar]] variance decomposition solution. Third, we introduce a suite of indices to assess the sensitivity of a predictor to model specification, advancing a series of sensitivity-adjusted contribution statistics that allow for more definite quantification of predictor relevance. Fourth, we describe the analytic efficiency of our proposed technique against the Budescu conditional dominance solution to the uneven contribution of predictors across all p! orderings.     

On evaluating the performance of different forecasters

In this article, we have evaluated the performance of different forecasters and tested association between their performances for different pairs of variables. We have used three data sets of track records of professional U.S. economic forecasters participating in the Blue Chip consensus forecasting service (the data sets contain the root mean square errors (RMSE) of different forecasters for different years). To evaluate the performance of forecasters we have covered three well-known tests, namely the usual F test (cf. Fisher (1923 Fisher, R. A., Mackenzie, M. A. (1923). Studied in crop variation II. The manurial response of different potato. Journal of Agricultural Science 13:311–320. [Google Scholar])), Kruskal Wallis test (cf. Kruskal and Wallis (1952 Kruskall, W. H., Wallis, W. A. (1952). Use of ranks in one-criterion variance analysis. Journal of American Statistical Association 47:583–621. [Google Scholar])), and Extension of Median test (cf. Daniel (1990 Daniel, W. W. (1990). Applied Nonparametric Statistics. Duxbury Classic Series. (2nd Ed.), Boston. [Google Scholar])). To test the association between the forecaster's performances for different pairs of variables, we have considered Gini mean correlation coefficient r_g1 (cf. Yitzhaki, S., and Olkin, I. (1991 Yitzhaki, S., Olkin, I. (1991). Concentration indices and concentration curves, in K. Mosler and M. Scarsini (eds.), Stochastic Orders and Decisions under Risk, Institute of Mathematical Statistics: Lecture-Notes Monograph Series, 19, 1991, 380–392. [Google Scholar]) and Yitzhaki (2003 Yitzaki, S. (2003). Gini mean difference: A superior measure of variability for non normal distribution. Metron-International Journal of Statistics, LXI:285–316. [Google Scholar])), Modified rank correlation coefficient (cf. Zimmerman (1994 Zimmerman, D. W. (1994). A Note on modified rank correlation. Journal of educational and Behavioral Statistics 19:357–362. [Google Scholar])) and three modifications of Spearman rank correlation coefficient. We have observed that different forecasters do not necessarily offer same average performance. Moreover, an evidence of association between two criteria does not always lead us reaching at the same decision. The outcomes of the study may help the practitioners in selecting the best forecaster(s) for policymaking purposes.     

Calibration approach estimation of the mean in stratified sampling and stratified double sampling

Calibration estimation improves the precision of the estimates of population parameters by incorporating specified auxiliary information. A class of calibration estimators has been proposed for estimating the population mean by making use of a set of calibration constraints in stratified sampling. The estimator of variance of the proposed calibration estimator of the mean is derived using a lower level calibration approach. The idea is extended for stratified double sampling. A simulation study is used to evaluate the performances of the proposed estimators by comparing them with the similar estimators developed by Tracy, Singh and Arnab (2003 Tracy, D.S., Singh, S., Arnab, R. (2003). Note on calibration in stratified and double sampling. Surv. Methodol. 29(1): 99–104. [Google Scholar]) based on different sets of calibration constraints.     

Edgeworth Expansion of the Moment-Based Test for Homogeneity in an NEF-QVF Mixture Model

In this article, we study the moment-based test procedure for a mixture distribution for the Natural exponential family with quadratic variance functions (NEF-QVF) family proposed by Ning et al. (2009b Ning, W., Zhang, S. G. and Yu, C. 2009b. A moment-based test for the homogeneity in mixture natural exponential family with quadratic variance functions. Statistical and Probability Letters, 79: 828–834. [Crossref] , [Google Scholar]) in the small sample size scenario. We derive the approximation for the null distribution of the test statistic by the Edgeworth expansion. The simulations are conducted for a binomial mixture distribution, which includes the situation corresponding to the detection of the linkage in the genetic analysis, with different sample sizes and family sizes at various significance levels. The simulation results show that our test performs reasonably well. We also apply the proposed method to the real clinical data to verify the significant difference between two drug treatments. The critical values associated with a binomial mixture distribution are also provided.     

Confidence Intervals for Linear Function of Variance Components in Two-Fold Nested Random Effects Model

In the article, we consider the unbalanced case of the two-way nested random effects model under partial balance. Using the method of generalized confidence intervals (GCIs) introduced in Weeranhandi (1993 Weeranhandi , S. ( 1993 ). Generalized confidence intervals . J. Amer. Statist. Assoc. 88 : 899 – 905 .[Taylor & Francis Online], [Web of Science ®] , [Google Scholar] 1995 Weeranhandi , S. ( 1995 ). Exact Statistical Methods for Data Analysis . New York : Springer-Verlag . [Google Scholar]), a new method is proposed for constructing confidence intervals on linear function of variance components. To compare the resulted intervals with the Modified Large Sample (MLS) intervals by Hernandez and Burdick (1993 Hernandez , R. P. , Burdick , R. K. ( 1993 ). Confidence intervals on the total variance in unbalanced two-fold nested designs . Biom. J. 35 : 515 – 522 .[Crossref] , [Google Scholar]), a simulation study is conducted. The results indicate that the proposed method performs better than the MLS method, especially for very unbalanced designs.     

A Regression Model for the AUC of Clustered Ordinal Test Results and Working Independent Optimal Weights

We study a regression model on the area under the receiver operating characteristic curves (AUC) for clustered (or repeatedly measured) test results. To account for cluster information, we consider a weighted estimating equation for Dodd and Pepe (2003 Dodd , L. , Pepe , M. ( 2003 ). Semiparametric regression for the area under the receiver operating charateristic curve . Journal of the American Statistical Association 98 : 409 – 417 .[Taylor & Francis Online], [Web of Science ®] , [Google Scholar])'s regression model with working independence weights. We find the optimal weight in the given class of working independence weights to minimize the variance (or MSE) of regression estimators. We apply the proposed procedure to analyzing our recent experiment on diagnosing a liver disorder. In this experiment, we investigated MRI images of patients having symptoms of potential liver disorder to compare the performance of different MRI picturing methods in testing for liver disorders.     

Empirical Likelihood in Causal Inference

This paper discusses the estimation of average treatment effects in observational causal inferences. By employing a working propensity score and two working regression models for treatment and control groups, Robins et al. (1994 Robins , J. M. , Rotnitzky , A. , Zhao , L. P. ( 1994 ). Estimation of regression coefficients when some regressors are not always observed . Journal of the American Statistical Association 89 : 846 – 866 .[Taylor &; Francis Online], [Web of Science ®] , [Google Scholar], 1995 Robins , J. M. , Rotnitzky , A. , Zhao , L. P. ( 1995 ). Analysis of semiparametric regression models for repeated outcomes in the presence of missing data . Journal of the American Statistical Association 90 : 106 – 121 .[Taylor &; Francis Online], [Web of Science ®] , [Google Scholar]) introduced the augmented inverse probability weighting (AIPW) method for estimation of average treatment effects, which extends the inverse probability weighting (IPW) method of Horvitz and Thompson (1952 Horvitz , D. G. , Thompson , D. J. ( 1952 ). A generalization of sampling without replacement from a finite universe . Journal of the American Statistical Association 47 : 663 – 685 .[Taylor &; Francis Online], [Web of Science ®] , [Google Scholar]); the AIPW estimators are locally efficient and doubly robust. In this paper, we study a hybrid of the empirical likelihood method and the method of moments by employing three estimating functions, which can generate estimators for average treatment effects that are locally efficient and doubly robust. The proposed estimators of average treatment effects are efficient for the given choice of three estimating functions when the working propensity score is correctly specified, and thus are more efficient than the AIPW estimators. In addition, we consider a regression method for estimation of the average treatment effects when working regression models for both the treatment and control groups are correctly specified; the asymptotic variance of the resulting estimator is no greater than the semiparametric variance bound characterized by the theory of Robins et al. (1994 Robins , J. M. , Rotnitzky , A. , Zhao , L. P. ( 1994 ). Estimation of regression coefficients when some regressors are not always observed . Journal of the American Statistical Association 89 : 846 – 866 .[Taylor &; Francis Online], [Web of Science ®] , [Google Scholar], 1995 Robins , J. M. , Rotnitzky , A. , Zhao , L. P. ( 1995 ). Analysis of semiparametric regression models for repeated outcomes in the presence of missing data . Journal of the American Statistical Association 90 : 106 – 121 .[Taylor &; Francis Online], [Web of Science ®] , [Google Scholar]). Finally, we present a simulation study to compare the finite-sample performance of various methods with respect to bias, efficiency, and robustness to model misspecification.