• Journal of Biomedical Engineering Research
• /
• v.19 no.2
• /
• pp.133-142
• /
• 1998

In Bayesian SPECT reconstruction, the incorporation of elaborate forms of priors can lead to improved quantitative performance in various statistical terms, such as bias and variance. In particular, the use of higher-order smoothing priors, such as the thin-plate prior, is known to exhibit improved bias behavior compared to the conventional smoothing priors such as the membrane prior. However, the bias advantage of the higher-order priors is effective only when the hyperparameters involved in the reconstruction algorithm are properly chosen. In this work, we further investigate the quantitative performance of the two representative smoothing priors-the thin plate and the membrane-by observing the behavior of the associated hyperparameters of the prior distributions. In our experiments we use Monte Carlo noise trials to calculate bias and variance of reconstruction estimates, and compare the performance of ML-EM estimates to that of regularized EM using both membrane and thin-plate priors, and also to that of filtered backprojection, where the membrane and thin plate models become simple apodizing filters of specified form. We finally show that the use of higher-order models yields excellent "robustness" in quantitative performance by demonstrating that the thin plate leads to very low bias error over a large range of hyperparameters, while keeping a reasonable variance. variance.

• Journal of Korean Institute of Industrial Engineers
• /
• v.38 no.1
• /
• pp.25-30
• /
• 2012

Desirability approaches have been proposed to find an optimum of multiple response problem. The existing desirability approaches use either of mean or min of individual desirability in aggregation of multiple responses. However, in order to find an optimum having high mean and low dispersion among individual desirability, the dispersion needs to be simultaneously considered with its mean. This study proposes bias and variance (BV) method which aggregates bias (ideal target-mean) and variance of individual desirability in multiple response optimization. The proposed BV method was applied to an example to evaluate its usefulness by comparing with existing methods. Evaluation results showed that the solution of BV method was a fairly good compared with DS (Derringer and Suich, 1980) and KL (Kim and Lin, 2000) methods. The BV method can be utilized to multiple response surface problems when decision makers want to find an optimum having high mean and low variance among responses.

• Communications for Statistical Applications and Methods
• /
• v.19 no.1
• /
• pp.23-32
• /
• 2012

This paper deals with the bias and variance of regression coefficient estimators in a finite population. We derive approximate formulas for the bias, variance and mean square error of two estimators when we select a fixed-size inclusion probability proportional to the size sample and then estimate regression coefficients by the ordinary least square estimator as well as the weighted least square estimator based on the selected sample data. Necessary and sufficient conditions for the comparison of the two estimators in terms of variance and mean square error are suggested. In addition, a simple example is introduced to numerically compare the variance and mean square error of the two estimators.

• Journal of the Korean Statistical Society
• /
• v.12 no.2
• /
• pp.81-90
• /
• 1983

The least squares estimator of disturbance variance in a regression model is biased under a serial correlation. Under the assumption of an AR(I), Theil(1971) crudely related the bias with the autocorrelation of the disturbances and the autocorrelation of the explanatory variable for a simple regression. In this paper we derive a relation which relates the bias with the autocorrelation of disturbances and the autocorrelation of explanatory variables for a multiple regression with improved precision.

• Journal of Korean Institute of Industrial Engineers
• /
• v.28 no.1
• /
• pp.112-127
• /
• 2002

A useful pattern is a pattern that contributes much to learning. For a classification problem those patterns near the class boundary surfaces carry more information to the classifier. For a regression problem the ones near the estimated surface carry more information. In both cases, the usefulness is defined only for those patterns either without error or with negligible error. Using only the useful patterns gives several benefits. First, computational complexity in memory and time for learning is decreased. Second, overfitting is avoided even when the learner is over-sized. Third, learning results in more stable learners. In this paper, we propose a pattern 'utility index' that measures the utility of an individual pattern. The utility index is based on the bias and variance of a pattern trained by a network ensemble. In classification, the pattern with a low bias and a high variance gets a high score. In regression, on the other hand, the one with a low bias and a low variance gets a high score. Based on the distribution of the utility index, the original training set is divided into a high-score group and a low-score group. Only the high-score group is then used for training. The proposed method is tested on synthetic and real-world benchmark datasets. The proposed approach gives a better or at least similar performance.

• Journal of Integrative Natural Science
• /
• v.5 no.4
• /
• pp.241-245
• /
• 2012

Categorical data collected based on complex sample design is not proper for the standard Pearson multinomial-based chi-squared test because the observations are not independent and identically distributed. This study investigates effects of bias of point estimator of population proportion and its variance estimator to the standard Pearson chi-squared test statistics when the sample is collected based on complex sampling scheme. This study examines the effect under two population homogeneity test. The standard Pearson test statistic can be partitioned into two parts; the first part is the weighted sum of ${\chi}^2_1$ with eigenvalues of design matrix as their weights, and the additional second part which is added due to the biases of the point estimator and its variance estimator. Our empirical analysis shows that even though the bias of point estimator is small, Pearson test statistic is very much inflated due to underestimate the variance of point estimator. In the connection of design-based variance estimator and its design matrix, the bigger the average of eigenvalues of design matrix is, the larger relative size of which the first component part to Pearson test statistic is taking.

• Proceedings of the Korean Statistical Society Conference
• /
• 2002.05a
• /
• pp.183-188
• /
• 2002

Multiple imputation, proposed by Rubin, is a procedure for handling missing data. One of the attractive parts of multiple imputation is the simplicity of the variance estimation formula. Because of the simplicity, it has been often abused and misused beyond its original prescription. This paper provides the bias of the multiple imputation variance estimator for a linear point estimator and discusses when the bias can be safely neglected.

• Journal of the Korean Statistical Society
• /
• v.24 no.2
• /
• pp.507-518
• /
• 1995

The ordinary least squares based estiamte $S^2$ of the disturbance variance is considered in the linear regression model when the disturbances follow the first-order moving-average process. It is shown that $S^2$ is weakly consistent estimate for the disturbance varaince without any restriction on the regressor matrix X. Also, simple exact bounds on the relative bias of $S^2$ are given in finite sample sizes.

• Proceedings of the Korean Operations and Management Science Society Conference
• /
• 2004.05a
• /
• pp.294-297
• /
• 2004

Mean squared error (MSE) is an effective criterion to combine the mean and the standard deviation responses in dual response surface optimization. The bias and variance components of MSE need to be weighted properly in the given problem situation. This paper proposes a systematic method to determine the relative weights of bias and variance in accordance with a decision maker's prior and posterior preference structure.

• Journal of the Korean Operations Research and Management Science Society
• /
• v.21 no.3
• /
• pp.159-174
• /
• 1996

Facing decision-making tasks, managers frequently make judgments, However, since managers are human beings, the fficiency of their judgments is limited. Two major sources of inefficiency in their judgments have been recognized : one is systematic deviations from normatively preferred decisions, so called bias or incorrect intuition, and the other is inconsistency in their judgments, i. e. erratic decision making variance. Rather than bias, variance is really expensive or damaging. Thus, if the inconsistency inmanagers judgments is removed, performance could be by far improved by virtue of the reduced random variance. One of the approaches to improve managerial judgment is to simply bring managers together by effectively moderating the random variance due to inconsistency. Focusing on combining judgments, this paper addresses many relevant issues such as why combining and how to combine judgments, and suggests methods and models to effectively aggregate subjective judgments, We conduct an experiment to validata the effectiveness of combining jugements over individual judgments. Various combining schemes are also evaluated in terms of their prective accuracy. Among them, mean bias based wighting scheme turns out the best. However, when available information is not enough to estimate the expertise of judges, simple and robust equal weighting might be more efficient and productive. This urges an imperative future research on the issue of how many and which ones to combine from a large set of experts.