• 응용통계연구
• /
• 제30권6호
• /
• pp.957-981
• /
• 2017

관측되지 않는 효과 또는 고정효과로 설명할 수 없는 분산 구조가 포함되어 정확한 모수 추정이 어려운 경우 체계적인 분석을 위해 일반화 선형 모형은 임의효과가 포함된 일반화 선형 혼합 모형으로 확장되었다. 본 연구에서는 일반화 선형 모형 중에서도 이분적인 반응변수를 다루는 로지스틱 회귀모형에 임의효과를 포함한 최대 우도 추정 방법을 설명한다. 그중에서도 라플라스 근사법, 가우스-에르미트 구적법, 적응 가우스-에르미트 구적법 그리고 유사가능도 우도에 대한 최대우도 추정법을 자세히 알아본다. 또한 제안한 방법을 사용하여 한국 복지 패널 데이터에서 정신건강과 생활만족도가 자원봉사활동에 미치는 영향에 대해 분석한다.

• Communications for Statistical Applications and Methods
• /
• 제23권4호
• /
• pp.321-334
• /
• 2016

In this paper, we propose power t distribution based on t distribution. We also study the properties of and inferences for power t model in order to solve the problem of real data showing both skewness and heavy tails. The comparison of skew t and power t distributions is based on density plots, skewness and kurtosis. Note that, at the given degree of freedom, the kurtosis's range of the power t model surpasses that of the skew t model at all times. We draw inferences for two parameters of the power t distribution and four parameters of the location-scale extension of power t distribution via maximum likelihood. The Fisher information matrix derived is nonsingular on the whole parametric space; in addition we obtain the profile log-likelihood functions on two parameters. The response plots for different sample sizes provide strong evidence for the estimators' existence and unicity. An application of the power t distribution suggests that the model can be very useful for real data.

• 대한기계학회논문집
• /
• 제18권8호
• /
• pp.2053-2061
• /
• 1994

The static and fatigue tensile tests have been conduted to predict the fatigue life of 8-harness satin woven and plain woven carbon/epoxy composite plates containing a circular hole. A fatigue residual strength degradation model, based on the assumption that the residual strength for unnotched specimen decreases monotonically, has been applied to predict statistically the fatigue life of materials used in this study. To determine the parameters(c, b and K) of the residual strength degradation model, the minimization technique and the maximum likelihood method are used. Agreement of the converted ultimate strength by using the minimization technique with the static ultimate strength is reasonably good. Therefore, the minimization technique is more adjustable in the determination of the parameter and the prediction of the fatigue life than the maximum likelihood method.

• Journal of the Korean Data and Information Science Society
• /
• 제26권2호
• /
• pp.313-322
• /
• 2015

본 연구에서는 하나의 균일분포 또는 퇴화분포와 두 개의 이항분포의 혼합분포 모형에 대하여 최우추정법을 소개하며, 제시된 모형에 대하여 시뮬레이션을 통해 최우추정량의 성질을 밝히며, 실험을 통해 얻은 강의 평가 자료에 대하여 퇴화분포를 가지는 혼합분포에 대하여 적용하여 보았다. 특히 퇴화분포는 한국의 문화 특성상 가운데 값을 선호하는 현상을 모형화하는데 유용하게 사용될 수 있음을 보였다.

• Communications for Statistical Applications and Methods
• /
• 제7권3호
• /
• pp.885-897
• /
• 2000

The use of an unweighted mean and of separate tests is part of the current practice for analyzing interlaboratory studies, and we hope to improve on this method. We fit, using maximum likelihood(ML), a rather intricate, multi-parameter measurement model with the material's true value as a latent variable in a situation where quite serviceable regression and ANOVA calculations have already been developed. The model fit leads to both a weighted estimate of he overall mean, and to tests for equality of means, slopes and variances. Maximum likelihood tests for difference among variances poses a challenge in that the likelihood can easily becoem unbounded. Thus the major objective become to provide a useful test of variance equality.

• Communications for Statistical Applications and Methods
• /
• 제20권3호
• /
• pp.217-223
• /
• 2013

Two influence diagnostic methods for the common mean model are proposed. First, an investigation of the influence of observations according to minor perturbations of the common mean model is made by adapting the local influence method which is based on the likelihood displacement. It is well known that the maximum likelihood estimates are in general sensitive to influential observations. Case-deletions can be a candidate for detecting influential observations. However, the maximum likelihood estimators are iteratively computed and therefore case-deletions involve an enormous amount of computations. An approximation by Newton's method to the maximum likelihood estimator obtained after a single observation was deleted can reduce much of computational burden, which will be treated in this work. A numerical example is given for illustration and it shows that the proposed diagnostic methods can be useful tools.

• Journal of the Korean Data and Information Science Society
• /
• 제10권2호
• /
• pp.465-472
• /
• 1999

We considered a bootstrap method for constructing confidenc intervals for a one parameter model using multinomial sampling. The convergence rates or the proposed bootstrap method are calculated for model-based maximum likelihood estimators(MLE) using multinomial sampling. Monte Carlo simulation was used to compare the performance of bootstrap methods with normal approximations in terms of the average coverage probability criterion.

• Journal of the Korean Statistical Society
• /
• 제24권1호
• /
• pp.1-18
• /
• 1995

The generalized logit model of nominal type with random regressors is studied for bootstrapping. In particular, asymptotic normality and consistency of bootstrap model estimators are derived. It is shown that the bootstrap approximation to the distribution of the maximum likelihood estimators is valid for alsomt all sample sequences.

• Journal of the Korean Statistical Society
• /
• 제27권4호
• /
• pp.495-506
• /
• 1998

The proposed class of generalized logistic models, indexed by an extra parameter, can be used to model or to examine symmetric or asymmetric discrepancies from the logistic model. When there are a finite number of different design points, we are mainly concerned with maximum likelihood estimation of parameters and in deriving their large sample behavior A score test and a bootstrap hypothesis test are also considered to check if the standard logistic model is appropriate to fit the data or if a generalization is needed .

• 음성과학
• /
• 제11권4호
• /
• pp.43-52
• /
• 2004

Model parameters in HMM based speech recognition systems are normally estimated using Maximum Likelihood Estimation(MLE). The MLE method is based mainly on the principle of statistical data fitting in terms of increasing the HMM likelihood. The optimality of this training criterion is conditioned on the availability of infinite amount of training data and the correct choice of model. However, in practice, neither of these conditions is satisfied. In this paper, we propose a training algorithm, MCE(Minimum Classification Error), to improve the performance of a speech recognizer detecting mispronunciation of a foreign language. During the conventional MLE(Maximum Likelihood Estimation) training, the model parameters are adjusted to increase the likelihood of the word strings corresponding to the training utterances without taking account of the probability of other possible word strings. In contrast to MLE, the MCE training scheme takes account of possible competing word hypotheses and tries to reduce the probability of incorrect hypotheses. The discriminant training method using MCE shows better recognition results than the MLE method does.