• The Korean Journal of Applied Statistics
• /
• v.16 no.2
• /
• pp.407-426
• /
• 2003

We consider the missing covariates problem in generalized estimating equations(GEE) model. If the covariate is partially missing, GEE can not be calculated. In this paper, we study the performance of 7 imputation methods to handle missing covariates in GEE models, and the properties of GEE estimators are investigated after missing covariates are imputed for ordinal data of repeated measurements. The 7 imputation methods include i) Naive Deletion ii) Sample Average Imputation iii) Row Average Imputation iv) Cross-wave Regression Imputation v) Carry-over Imputation vi) Bayesian Bootstrap vii) Approximate Bayesian Bootstrap. A Monte-Carlo simulation is used to compare the performance of these methods. For the missing mechanism generating the missing data, we assume ignorable nonresponse. Furthermore, we generate missing covariates with or without considering wave nonresp onse patterns.

• The Korean Journal of Applied Statistics
• /
• v.25 no.1
• /
• pp.29-43
• /
• 2012

Financial data such as stock index returns and exchange rates have the properties of heavy tail and asymmetry compared to normal distribution. When we estimate VaR using the GARCH model (with the conditional return distribution of normal) it shows the tendency of the lower estimation and clustering in the losses over the estimated VaR. In this paper, we argue that this problem can be resolved through the adaptation of the unbounded Johnson distribution as that of the condition return. We also compare this model with the GARCH with the conditional return distribution of normal and student-t. Using the losses exceed the ex-ante VaR, estimates, we check the validity of the GARCH models through the failure proportion test and the clustering test. We nd that the GARCH model with conditional return distribution of unbounded Johnson provides an appropriate estimation of the VaR and does not occur the clustering of violations.

• The Korean Journal of Applied Statistics
• /
• v.20 no.1
• /
• pp.53-60
• /
• 2007

Conditionally heteroscedastic time series models such as GARCH processes have frequently provided useful approximations to the real aspects of financial time series. It is not uncommon that financial time series exhibits near non-stationary, say, integrated phenomenon. For stationary GARCH processes, a shock to the current conditional variance will be exponentially converging to zero and thus asymptotically negligible for the future conditional variance. However, for the case of integrated process, the effect will remain for a long time, i.e., we have a persistent effect of a current shock on the future observations. We are here concerned with providing empirical evidences of persistent GARCH(1,1) for various fifteen domestic financial time series including KOSPI, KOSDAQ and won-dollar exchange rate. To this end, kurtosis and Integrated-GARCH(1,1) fits are reported for each data.

• The Korean Journal of Applied Statistics
• /
• v.21 no.6
• /
• pp.1045-1063
• /
• 2008

This paper derives conversion formulas from yearly-based absolute rates of decrements to monthly-based rates of decrement due to cause j under FI (fractional age independence) assumption that is a generalization of UDD assumption. Next, it suggests conversion formulas from monthly-based absoluterates of decrements to monthly-based rates of decrement due to cause j under FI assumption. In addition, it calculates conversion formulas from yearly-based rates of decrement due to cause j to the corresponding monthly-based absolute rates of decrements under FI assumption. Some numerical examples are discussed.

• The Korean Journal of Applied Statistics
• /
• v.21 no.6
• /
• pp.923-932
• /
• 2008

We are likely to face complex multivariate data which can be characterized by having a non-trivial correlation structure. For instance, omitted covariates may simultaneously affect more than one count in clustered data; hence, the modeling of the correlation structure is important for the efficiency of the estimator and the computation of correct standard errors, i.e., valid inference. A standard way to insert dependence among counts is to assume that they share some common unobservable variables. For this assumption, we fitted correlated random effect models considering multilevel model. Estimation was carried out by adopting the semiparametric approach through a finite mixture EM algorithm without parametric assumptions upon the random coefficients distribution.

• The Korean Journal of Applied Statistics
• /
• v.26 no.3
• /
• pp.453-470
• /
• 2013

In this paper, we consider a Bayesian analysis of the dose-effect relationship of cadmium to evaluate a benchmark dose(BMD). For this purpose, two dose-response curves commonly used in the toxicity study are fitted based on Bayesian methods to the data collected from the scientific literature on cadmium toxicity. Specifically, Bayesian meta-analysis and hierarchical modeling build an overall dose-effect relationship that use a piecewise linear model and Hill model, where the inter-study heterogeneity and inter-individual variability of dose and effect such as gender, age and ethnicity are accounted. Estimation of the unknown parameters is made by using a Markov chain Monte Carlo algorithm based user-friendly software WinBUGS. Benchmark dose estimates are evaluated for various cut-offs and compared with different tested subpopulations with with gender, age and ethnicity based on these two Bayesian hierarchical models.

• The Korean Journal of Applied Statistics
• /
• v.30 no.6
• /
• pp.1015-1025
• /
• 2017

Bayesian personalized ranking (BPR) is a state-of-the-art recommendation system techniques for implicit feedback data. Unfortunately, there might be a loss of information because the BPR model considers only the binary transformation of implicit feedback that is non-binary data in most cases. We propose a modified BPR method using a level of confidence based on the size or strength of implicit feedback to overcome this limitation. The proposed method is useful because it still has a structure of interpretable models for underlying personalized ranking i.e., personal pairwise preferences as in the BPR and that it is capable to reflect a numerical size or the strength of implicit feedback. We propose a computation algorithm based on stochastic gradient descent for the numerical implementation of our proposal. Furthermore, we also show the usefulness of our proposed method compared to ordinary BPR via an analysis of steam video games data.

• The Korean Journal of Applied Statistics
• /
• v.30 no.6
• /
• pp.901-916
• /
• 2017

Recently, the industrial paradigm shift to the fourth industry has already begun, and the importance of patents as intangible intellectual property in the fourth industry era is increasing day by day. Since the technical valuation of a patent is calculated according to the opinion of experts, it is costly and time consuming, and hence, the quality of the patent is judged based on subjective opinions of non-experts. Therefore, it is necessary to develop an objective and rational evaluation system for the qualitative level of patents. In this paper, we classify the valuation of patents into technicality, rights, and usability, and consider the quantitative and objective evaluation modeling of patents using Bayesian structural equation model. In particular, based on the data collected by the Korea Invention Promotion Association, we apply the Bayesian approach, which is capable of stable modeling even under small samples by using prior information, and the structural equation model, which is excellent for modeling and evaluating qualitative performance that is difficult to measure directly, to develop a patent evaluation model.

• The Korean Journal of Applied Statistics
• /
• v.32 no.6
• /
• pp.889-898
• /
• 2019

The problem of missing value imputation for variables in surveys that include item missing becomes complicated if outliers and logical boundary conditions between other survey items cannot be ignored. If there are outliers and boundaries in a variable including missing values, imputed values based on previous regression-based imputation methods are likely to be biased and not meet boundary conditions. In this paper, we approach these difficulties in imputation by combining various robust regression models and multiple imputation methods. Through a simulation study on various scenarios of outliers and boundaries, we find and discuss the optimal combination of robust regression and multiple imputation method.

• The Korean Journal of Applied Statistics
• /
• v.32 no.6
• /
• pp.899-908
• /
• 2019

Massive in-depth neural networks with numerous parameters are powerful machine learning methods, but they have overfitting problems due to the excessive flexibility of the models. Dropout is one methods to overcome the problem of oversized neural networks. It is also an effective method that randomly drops input and hidden nodes from the neural network during training. Every sample is fed to a thinned network from an exponential number of different networks. In this study, instead of feeding one sample for each thinned network, two or more samples are used in fitting for one thinned network known as a Hybrid Dropout. Simulation results using real data show that the new method improves the stability of estimates and reduces the minimum error for the verification data.