• Proceedings of the Korean Statistical Society Conference
• /
• 2003.05a
• /
• pp.127-130
• /
• 2003

주변분포가 Laplace 분포인 세 가지 형태의 이변량 Laplace 분포를 연구한다. 각각의 이변량 Laplace 분포의 확률밀도함수와 누적분포함수를 유도하고, 분포의 그래프를 그려봄으로써 분포의 형태를 알아본다. 조건부 적률을 정리하여 조건부 첨도와 조건부 왜도를 구하고 분포의 성질을 파악한다. 상관계수를 구하여 다른 이변량 분포의 상관계수와 비교해 보았다. 그리고 정의된 분포함수를 응용하여 이변량 Laplace 분포를 따르는 난수벡터를 발생하는 알고리즘을 제안하였으며, 생성된 난수벡터의 표본으로부터 구한 표본평균과 중앙값의 분산-공분산 행렬식을 구하고 이변량 정규분포에 대응하는 행렬식과 비교 토론하였다.

• Journal of KIISE
• /
• v.41 no.10
• /
• pp.799-809
• /
• 2014

In this paper, we address the problem of computing Pearson correlation coefficients and Spearman's rank correlation coefficients in a secure manner while data providers preserve privacy of their own data in distributed environment. For a data mining or data analysis in the distributed environment, data providers(data owners) need to share their original data with each other. However, the original data may often contain very sensitive information, and thus, data providers do not prefer to disclose their original data for preserving privacy. In this paper, we formally define the secure correlation computation, SCC in short, as the problem of computing correlation coefficients in the distributed computing environment while preserving the data privacy (i.e., not disclosing the sensitive data) of multiple data providers. We then present SCC solutions for Pearson and Spearman's correlation coefficients using secure scalar product. We show the correctness and secure property of the proposed solutions by presenting theorems and proving them formally. We also empirically show that the proposed solutions can be used for practical applications in the performance aspect.

• The Korean Journal of Applied Statistics
• /
• v.28 no.3
• /
• pp.385-392
• /
• 2015

Principal component analysis (PCA), a dimension-reduction technique, is usually implemented after the variables are standardized when the measurement unit of variables are different. To standardize a variable we divide it by its standard deviation. But there is another way to transform a variable to be independent of its measurement unit. It is to divide it by its mean rather than standard deviation. Implementing PCA on standardized variables is equivalent to implementing PCA with a correlation matrix of original variables. Similarly, implementing PCA on the transformed variables divided by their means is equivalent to implementing PCA with a matrix related to the coefficients of variation of the original variables. We explain why we need to implement PCA on the variables transformed by their means.

• The Korean Journal of Applied Statistics
• /
• v.37 no.6
• /
• pp.783-799
• /
• 2024

Longitudinal data are measured repeatedly over time from the same subject. Therefore, the repeated outcomes have correlations, and it is necessary to estimate the covariate effect on the response variable while explaining the correlations. In longitudinal ordinal data analysis, the covariate effect is estimated using generalized linear mixed models using a logit link function or a probit link function. In this paper, we review the generalized linear mixed models and marginalized models with the two types of link functions for longitudinal ordinal data analysis. Specifically, a Bayesian cumulative probit linear mixed model with the probit link function is used to analyze Korean workplace panel survey (WPS) data, which is longitudinal ordinal data. In the model, the correlation matrix is high-dimensional and positive definite, and it is estimated using the hypersphere decomposition. In the WPS data, corporate training participation rate is considered as a response variable. Assuming different correlation structures, several models are compared. For the most suitable model, some explanatory variables, the annual effect, profit sharing schemes status, average annual training hours per person, and labor union status, have effects on corporate training participation rate.

• The Korean Journal of Applied Statistics
• /
• v.26 no.1
• /
• pp.141-149
• /
• 2013

We present methods for studying the log-density ratio that enables the selection of the predictors and the form to be included in the logistic regression model. Under bivariate normal distributional assumptions, we investigate the form of the log-density ratio as a function of two predictors. If two covariance matrices are equal, then the crossproduct and quadratic terms are not needed. If the variables are uncorrelated, we do not need the crossproduct terms, but we still need the linear and quadratic terms. We also explore other conditions in which the crossproduct and quadratic terms are not needed in the logistic regression model.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2009.10a
• /
• pp.471-473
• /
• 2009

In this paper, we propose a double talk and noise robust acoustic echo canceler for hands-free telephony applications. The proposed system includes a double-talk detection method that detects the double-talk durations, which uses covariance between microphone input signa and estimated microphone input signal. And proposed adaptive algorithm for estimating acoustic echo path, uses normalized auto-covariance matrix of input signal with multiplication of residual error power and projection order of AP(affine projeciton) algorithm. It is confirmed that the proposed algorithm shows better performance from acoustic interference cancellation (AIC) viewpoint in double talk and noisy environments.

• The Korean Journal of Applied Statistics
• /
• v.28 no.2
• /
• pp.201-210
• /
• 2015

Repeated measurements on each variables of interest often arise in bioscience or medical research. We need to account for correlations among repeated measurements to assess the correlation between two variables in the presence of replication. This paper reviews methods to estimate a correlation coefficient between two variables in repeated measurements using the variance-covariance matrix of linear mixed effect models. We analyze acoustic radiation force impulse imaging (ARFI) data to assess correlation between three shear wave velocity (SWV) measurements in liver or spleen and spleen length by ultrasonography. We present how to obtain parameter estimates for the variance-covariance matrix and correlations in mixed effects models using PROC MIXED in SAS.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.23 no.8
• /
• pp.2072-2079
• /
• 1998

In this paper, we propose a new decorrelating detector called pseudo-inverse-filtering type decorrelating detector for asynchronous CDMA channels. We first show that the matched filtering and decorrelating operations of the existing decorrelating detectors can be replaced with the pseudo-inverse filtering operations in synchronous channels, and using this fact we show that the decorrelating detector has the largest SNR among the linear detectors that can eliminate MAI. Then we introduce asynchronous pseudo-inverse filtering type decorrelating detector by extending this result for asynchronous channels, and discuss implementation methods of the proposed decorrelating detectors. Since the proposed scheme employs a decentralized structure for updating coefficients, it has the flexibility to add/remove users. Finally we analyze the performance of the proposed decorrelating detector in terms of the bit error rate, and examine its performance improvements over the conventional detectors through computer simulations.

• Journal of Animal Science and Technology
• /
• v.55 no.3
• /
• pp.165-171
• /
• 2013

Genomic relationship matrix (GRM) was constructed using whole genome SNP markers of swine and genomic breeding value was estimated by substitution of the numerator relationship matrix (NRM) based on pedigree information to GRM. Genotypes of 40,706 SNP markers from 448 pure Landrace pigs were used in this study and five kinds of GRM construction methods, G05, GMF, GOF, $GOF^*$ and GN, were compared with each other and with NRM. Coefficients of GOF considering each of observed allele frequencies showed the lowest deviation with coefficients of NRM and as coefficients of GMF considering the average minor allele frequency showed huge deviation from coefficients of NRM, movement of mean was expected by methods of allele frequency consideration. All GRM construction methods, except for $GOF^*$, showed normally distributed Mendelian sampling. As the result of breeding value (BV) estimation for days to 90 kg (D90KG) and average back-fat thickness (ABF) using NRM and GRM, correlation between BV of NRM and GRM was the highest by GOF and as genetic variance was overestimated by $GOF^*$, it was confirmed that scale of GRM is closely related with estimation of genetic variance. With the same amount of phenotype information, accuracy of BV based on genomic information was higher than BV based on pedigree information and these symptoms were more obvious for ABF then D90KG. Genetic evaluation of animal using relationship matrix by genomic information could be useful when there is lack of phenotype or relationship and prediction of BV for young animals without phenotype.