• Journal of Korea Water Resources Association
• /
• v.45 no.2
• /
• pp.191-201
• /
• 2012

This study reviewed the parameter estimation procedure of the Freund bivariate exponential distribution for the decision of the annual maximum rainfall event. The method of moments was reviewed first, whose results were compared with those from the method of maximum likelihood. Both methods were applied to the hourly rainfall data of the Seoul rain gauge station measured from 1961 to 2010 to select the annual maximum rainfall events, which were also compared each other. The results derived are as follows. First, when applying the method of moments for the parameter estimation, it was found necessary to consider the correlation coefficient between the two variables as well as the mean and variance. Second, the method of maximum likelihood was better to reproduce the mean, but the method of moments was better to reproduce the annual variation of the variance. Third, The annual maximum rainfall events derived were very similar in both cases. Among differently selected annual maximum rainfall events, those with the higher rainfall amount were selected by the method of maximum likelihood, but those with the higher rainfall intensity by the method of moments.

• The Korean Journal of Applied Statistics
• /
• v.35 no.4
• /
• pp.501-515
• /
• 2022

Despite the stylized statistical features of returns of financial returns such as fat-tailed distribution and leverage effect, no stochastic volatility models that can explicitly capture these features have been presented in the existing frequentist approach. we propose an approximate parameterization of stochastic volatility models that can explicitly capture the fat-tailed distribution and leverage effect of financial returns and a maximum likelihood estimation of the model using Langrock et al. (2012)'s hidden Markov model approximation in a frequentist approach. Through extensive simulation experiments and an empirical analysis, we present the statistical evidences validating the efficacy and accuracy of proposed parameterization.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2016.05a
• /
• pp.280-280
• /
• 2016

특정 자료의 시간의 흐름에 따른 예측치를 추정하는 방법으로 AR Model 즉, 자기회귀모형이 많이 사용되고 있다. AR Model은 변수의 현재 값을 과거 값의 함수로 나타내게 되는데, 이런 시계열 분석 모델을 사용할 때 매개변수의 추정 과정이 필수적으로 요구된다. 일반적으로 매개변수를 추정하는 방법에는 확률적근사법(stochastic approximation), 최소제곱법(method of least square), 자기상관법(method of autocorrelation method), 최우도법(method of maximum likelihood) 등이 있다. AR Model에서 가장 많이 사용되는 최우도법은 표본크기가 충분히 클 때 가장 효율적인 방법으로 평가되지만 수치적으로 해를 구하는 과정이 복잡한 경우가 많으며, 해를 구하지 못하는 어려움이 따르기도 한다. 또한 표본 크기가 작을 때 일반적으로 잘 일치하지 않은 결과를 얻게 된다. 우리나라의 강우, 유량 등의 자료는 자료의 수가 적은 경우가 많기 때문에 최우도법을 통한 매개변수 추정 시 불확실성이 내재되어있지만 그것을 정량적으로 제시하는데 한계가 있다. 본 연구에서는 AR Model의 매개변수 추정 시 Bayesian 기법으로 매개변수의 사후분포(posterior distribution)를 제공하여 매개변수의 불확실성 구간을 정량적으로 표현하게 됨으로써, 시계열 분석을 통해 보다 신뢰성 있는 예측치를 얻을 수 있으리라 판단된다.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2009.05a
• /
• pp.1272-1276
• /
• 2009

극치강우사상에 의한 설계 홍수량의 갑작스런 증 감은 홍수, 가뭄과 같은 기상학적 요인에 기인한 재난을 발생시킨다. 많은 연구자들은 보다 정확한 확률강우량의 예측과 유출량의 예측을 위해 많은 노력을 하고 있다. 본 연구에서는 강원도 강릉 강우관측소를 대상으로 강우-빈도곡선의 불확실성 분석을 수행하였다. 관측 자료의 수집에서 발생하는 불확실성을 최소화 하고자 ARMA 모형을 이용하여 합성강우자료를 구축하였으며, 발생된 합성강우량을 Bootstrap 방법을 이용하여 대규모의 자료집단으로 발생시킴으로서 신뢰구간에 사용할 자료집단을 발생시켰다. 본 연구에서는 극치강우사상에 적합한 것으로 알려진 Gumbel 분포와 일반극치 분포(GEV 분포) 모형을 선정하였으며 각 확률분포모형에 대한 매개변수 추정방법으로 최우도법, 확률가중모멘트법 그리고 베이지안 추론방법을 사용하여 각 매개변수의 최후 추정치를 산정하였다. 또한 원 자료를 이용하여 최우도법, 확률가중모멘트법 그리고 베이지안 추론방법을 통해 매개변수를 산정 후 강우-빈도 곡선을 추정하여 합성강우자료의 Bootstrap 방법에 의해 발생된 자료로부터 산정한 강우-빈도 곡선의 신뢰구간과 비교함으로서 불확실성이 낮은 확률강우량을 산정할 수 있는 매개변수 추정방법을 평가하고자하였다.

• Journal of the Korean Data and Information Science Society
• /
• v.24 no.4
• /
• pp.755-762
• /
• 2013

In this paper we consider the NHPP software reliability model. And we deal with the maximum likelihood estimation and the Bayesian estimation with conjugate prior for parameter inference in the mean value function of Goel-Okumoto model (1979). The parameter estimates for the proposed model is presented by MLE and Bayes estimator in data set. We compare the predicted number of faults with the actual data set using the proposed mean value function.

• Journal of KIISE:Software and Applications
• /
• v.28 no.1
• /
• pp.46-51
• /
• 2001

2D로 확장한 HMM은 다수 제안되었지만 엄밀한 의미에 있어서 2D HMM이라고 하기에 부족한 점이 많다. 본 논문에서는 기존의 랜덤 필드 모형이 아닌 새로운 2D HMM을 제안한다. 상하 및 좌우 방향의 causal chain 관계를 가정하고 완전한 격자 형성 조건을 두어 2D HMM의 평가, 매개 변수를 추정하는 알고리즘을 제시하였다. 각각의 알고리즘은 동적 프로그래밍과 최우 추정법에 근거한 것이다. 변수 추정 알고리즘은 반복적으로 이루어지며 국소 최적치에 수렴함을 보였다.

• Journal of Korea Water Resources Association
• /
• v.43 no.2
• /
• pp.187-199
• /
• 2010

This study evaluated applicability and confidence of probability rainfalls estimated by the non-stationary rainfall frequency analysis which was recently developed. Using rainfall data at 4 sites which have an obvious increasing trend in observations, we estimated 3 type probability rainfalls; probability rainfalls from stationary rainfall frequency analysis using data from 1973-1997, probability rainfalls from stationary rainfall frequency analysis using data from 1973-2006, probability rainfalls from non-stationary rainfall frequency analysis assuming that the current year is 1997 and the target year is 2006. Based on the comparison of residuals from 3 probability rainfalls, the non-stationary rainfall frequency analysis provided more effective and well-directed estimates of probability rainfalls in the target year. Using Bootstrap resampling, this study also evaluated the parameter estimation methods for the non-stationary rainfall frequency analysis based on confidence intervals. The confidence interval length estimated by the maximum likelihood estimation (MLE) is narrower than the probability weighted moments (PWM). The results indicated that MLE provides more proper confidence than PWM for non-stationary probability rainfalls.

• KSCE Journal of Civil and Environmental Engineering Research
• /
• v.13 no.4
• /
• pp.151-161
• /
• 1993

The log-Gumbel distribution in real space is defined by transforming the conventional log-Gumbel distribution in log space. For this model, the parameter estimation techniques are applied based on the methods of moments, maximum likelihood and probability weighted moments. The asymptotic variances of estimator of the quantiles for each estimation method are derived to find the confidence limits for a given return period. Finally, the log-Gumbel model is applied to actual flood data to estimate the parameters, quantiles and confidence limits.

• Journal of Korea Water Resources Association
• /
• v.39 no.8 s.169
• /
• pp.659-668
• /
• 2006

The uncertainty of the risk of failure of hydraulic structures can be determined by estimating the variance of the risk of failure based on the methods of moments, probability weighted moments, and maximum likelihood assuming that the underlying model is the Gumbel distribution. In this paper, the variance of the risk of failure was derived. Monte Carlo simulation was peformed to verify the characteristics of the derived formulas for various sample size, design life, nonexceedance probability, and variation coefficient. As the results, PWM showed the smallest relative bias and root mean square error than the others while ML showed the smallest ones for relatively large sample siBes regardless of design life and nonexceedance probability. Also, it was found that variation coefficient does not effect on the relative bias and relative root mean square error.

• The Korean Journal of Applied Statistics
• /
• v.5 no.2
• /
• pp.181-192
• /
• 1992

We consider the estimation of the regression surface in generalized linear models based on tensor-product B-splines in a data-dependent way. Our approach is to use maximum likelihood method to estimate the regression function by a function from a space of tensor-product B-splines that have a finite number of knots and are linear in the tails. The knots are placed at selected order statistics of each coordinate of the sample data. The number of knots is determined by minimizing a variant of AIC. A numerical example is used to illustrate the performance of the tensor spline estimates.