• Communications for Statistical Applications and Methods
• /
• v.25 no.5
• /
• pp.453-470
• /
• 2018

We study how to distinguish the parameters of the sparse autoregressive (AR) process from zero using a non-convex penalized estimation. A class of non-convex penalties are considered that include the smoothly clipped absolute deviation and minimax concave penalties as special examples. We prove that the penalized estimators achieve some standard theoretical properties such as weak and strong oracle properties which have been proved in sparse linear regression framework. The results hold when the maximal order of the AR process increases to infinity and the minimal size of true non-zero parameters decreases toward zero as the sample size increases. Further, we construct a practical method to select tuning parameters using generalized information criterion, of which the minimizer asymptotically recovers the best theoretical non-penalized estimator of the sparse AR process. Simulation studies are given to confirm the theoretical results.

• Atmosphere
• /
• v.18 no.2
• /
• pp.111-120
• /
• 2008

This study presents a change-point in the 30 years (1976-2005) time series of the annual and the heavy precipitation characteristics (amount, days and intensity) averaged over South Korea using Bayesian approach. The criterion for the heavy precipitation used in this study is 80 mm/day. Using non-informative priors, the exact Bayes estimators of parameters and unknown change-point are obtained. Also, the posterior probability and 90% highest posterior density credible intervals for the mean differences between before and after the change-point are examined. The results show that a single change-point in the precipitation intensity and the heavy precipitation characteristics has occurred around 1996. As the results, the precipitation intensity and heavy precipitation characteristics have clearly increased after the change-point. However, the annual precipitation amount and days show a statistically insignificant single change-point model. These results are consistent with earlier works based on a simple linear regression model.

• The Korean Journal of Applied Statistics
• /
• v.9 no.1
• /
• pp.153-164
• /
• 1996

We discussed estimation of parameters using Gibbs sampler under order restriction on the parameters. Two well-knwon probability models, ordered exponential family and binomial distribution, are considered. We derived full conditional distributions(FCD) and also used one-for-one sampling algorithm to sample from the FCD's under order restrictions. Finally through two real data sets we compared three kinds of estimators; isotonic regression estimator, isotonic Bayesian estimator and the estimator using Gibbs sampler.

• Journal of the Korean Data and Information Science Society
• /
• v.9 no.1
• /
• pp.29-36
• /
• 1998

A Simple, closed form expression is derived for a linear smoothing spline by Eubank (1994, 1997). We introduced his estimater and studied how well examples are fitted to this estimator with the values of minimum generalized cross validation.

• Journal of Korean Institute of Industrial Engineers
• /
• v.25 no.4
• /
• pp.510-516
• /
• 1999

This paper proposes methods of estimating lifetime distribution from incomplete field data under parametric regression models. Failure-record data-failure times and covariates-reported to the manufacturer can be seriously incomplete for satisfactory inference since only reported failures are recorded. This paper assumes that within-warranty data are reported with probability $P_1$ ($\leq1$) and after-warranty data are reported with Methods of obtaining pseudo and after-warranty data are reported with $P_2$ (< $P_1$). Methods of obtaining pseudo maximum likelihood estimators(PMLEs) are outlined, their asymptotic properties are studied, and specific formulas for Weibull distribution are obtained. Simulation studies are perfumed to investigate the effects of follow-up percentage on the PMLEs.

• The Korean Journal of Applied Statistics
• /
• v.6 no.2
• /
• pp.237-245
• /
• 1993

A nonparametric test for the parallelisim of k regression lines against ordered alternatives is proposed. The test statistic is weighted Jonckheere-type statistic applied to slope estimators obtained from each lines. The distribution of the proposed test statistic is asymptotically distribution-free. From the viewpoint of efficiencies, the proposed test desirable properties and is more efficient than the other nonparametric tests.

• The Korean Journal of Applied Statistics
• /
• v.5 no.2
• /
• pp.283-292
• /
• 1992

Updating algorithms are studied for the basic statistics (mean, variance). For a linear model, a recursive formulae for least squares estimators of regression coefficients, residual sum of squares and variance-covariance matrix are also studied. Hotelling's $T^2$ statistics can be calculated recursively using the recursive formulae of mean vector and variance-covariance matrix without computing the sample variance-covariance matrix at each stage.

• The Korean Journal of Applied Statistics
• /
• v.25 no.6
• /
• pp.1037-1047
• /
• 2012

The generalized linear mixed model(GLMM) is widely used in fitting categorical responses of clustered data. In the numerical approximation of likelihood function the normality is assumed for the random effects distribution; subsequently, the commercial statistical packages also routinely fit GLMM under this normality assumption. We may also encounter departures from the distributional assumption on the response variable. It would be interesting to investigate the impact on the estimates of parameters under misspecification of distributions; however, there has been limited researche on these topics. We study the sensitivity or robustness of the maximum likelihood estimators(MLEs) of GLMM for counts data when the true underlying distribution is normal, gamma, exponential, and a mixture of two normal distributions. We also consider the effects on the MLEs when we fit Poisson-normal GLMM whereas the outcomes are generated from the negative binomial distribution with overdispersion. Through a small scale Monte Carlo study we check the empirical coverage probabilities of parameters and biases of MLEs of GLMM.

• International Journal of Quality Innovation
• /
• v.4 no.2
• /
• pp.16-31
• /
• 2003

A reducing variation of quality characteristics is a typical example of quality improvement. In such a case, we treat the quality characteristic, as a response variable and need to find active factors affecting the response from many candidate factors since reducing the variation of the response will be achieved by reducing variation of the active factors. In this paper, we first derive a method of selecting an active factor by linear regression. It is well known that correlation between factors deteriorates the precision of estimators. We, therefore, examine robustness of the selecting method against the correlation in the data set and derive an evaluation method of the deterioration brought by the correlation. Furthermore, some examples of selecting and evaluation methods are shown to demonstrate practical usage of the methods.

• International conference on construction engineering and project management
• /
• 2009.05a
• /
• pp.417-424
• /
• 2009

For a construction project to progress smoothly, effective cost estimation is vital, particularly in the conceptual and schematic design stages. In these early phases, despite the fact that initial estimates are highly sensitive to changes in project scope, owners require accurate forecasts which reflect their supplying information. Thus, cost estimators need effective estimation strategies. Practically, parametric cost estimates are the most commonly used method in these initial phases, which utilizes historical cost data (Karshenas 1984, Kirkham 2007). Hence, compilation of historical data regarding appropriate cost variance governing parameters is a prime requirement. However, precedent practice of data mining (data preprocessing) for denoising internal errors or abnormal values is needed before compilation. As an effort to deal with this issue, this research proposed a statistical methodology for data preprocessing and verified that data preprocessing has a positive impact on the enhancement of estimate accuracy and stability. Moreover, Statistically Preprocessed data Based Parametric (SPBP) cost models are developed based on multiple regression equations and verified their effectiveness compared with conventional cost models.