SNS에서의 소비자 정보 확산에 있어 사회적 전염과 랜덤효과에 관한 탐색적 연구

한상만,옥경영 한국소비자학회 2012 소비자학연구 Vol.23 No.2

본 연구는 사회적 네트워크(SNS, Social Networking Site)에서 이루어지는 정보 확산에 있어 연결되어 있는 이웃에 의한 사회적 전염(social contagion)뿐 아니라 랜덤효과(random effects)의 영향을 함께 파악하고자 사회 적 폭포현상(social cascade)과 최대크기콤포넌트(largest component)라는 2가지 지표를 사용하여 이를 분석하 였고, 또한 이를 랜덤그래프와 비교하여 랜덤효과(random effects)의 영향을 살펴보았다. 여기서 사회적 전염이란 사회적 네트워크에서 연결된 소비자들이 이웃의 영향을 받아 정보를 수용함으로써 정보 확산이 이루어지는 것이며, 랜덤효과(random effects)란 사회적 전염에 의하지 않고 스스로 정보를 찾거나 매스미디어의 영향에 의해 정보를 수용함으로써 정보 확산이 이루어지는 것을 말한다. 본 연구는 우리나라 SNS에서 수집된 데이터에서 도출한 실제 사회적 네트워크를 대상으로 분석하였다. 그 결과, 사회적 네트워크에서 발생하는 정보 확산에 있어 기존에 알려진 사회적 전염에 의한 확산뿐 아니라 랜덤효과에 의한 확산이 중요한 것으로 나타났다. 즉, 사회적 네트워크에서의 성 공적인 정보 확산에 있어 이웃의 영향에 의해 발생하는 사회적 전염과 매스미디어 등의 랜덤효과가 함께 존재해야 함을 발견하였고, 특히 대규모 확산의 경우 랜덤효과에 의한 확산이 차지하는 중요성이 더 커짐을 발견한 것이 이 논문의 공헌점이다. In this study, we explore the role of social contagion and random effects in consumer information diffusion. We adopted a social cascading index and the largest component size as the measures of social contagion. Social contagion is defined as the adoption of information influenced by his/her neighbors, while random effects are defined as the adoption of information not influenced by his/her neighbors. The random effects include the influence of Mass media on the information adoption. The results show that there exist two types of information diffusion patterns: the first pattern shows a peak of diffusion at the very early stage followed by a sudden decrease of diffusion. The second type shows a continuous increase in diffusion until it reaches the peak in the middle stage of the diffusion path. We found that in the case of the first pattern of diffusion with a peak in the very early stage followed by a sudden decrease, the adoption of information is coming mostly from the social contagion mechanism. In the second type of diffusion with a continuous increase until its peak in the middle stage of diffusion path, the adoption of information is not coming just from the social contagion mechanism, but also from the random effects mechanism. The contribution of our paper is two fold: First, we have suggested a dual mechanism of consumer information adoption where consumers adopt new information through both social contagion and random effects. Second, we found that in order to generate a successful information diffusion, social contagion must be combined with the random effects. Social contagion alone cannot create a huge successful diffusion. A successful information diffusion always needs not only a successful social contagion but also "random effects."

Prediction of random effects with misspecified AFT random effect models

Lin Hao,하일도 한국데이터정보과학회 2022 한국데이터정보과학회지 Vol.33 No.5

The accelerated failure time (AFT) model with random effects has been widely used for clustered or correlated time-to-event data as an alternative to frailty model which is the Cox's proportional hazards model with random effects. The AFT random effect model usually assumes a normal distribution for random effect distribution. It is well known that the estimated regression parameters in the AFT model are robust against various violations of the assumed model. However, the impact of prediction (or estimation) of random effect, when the assumed normal random effect is misspecified, has been relatively less studied. In this paper, we investigate the impact of misspecification of normal random effect distribution on the prediction of random effect under the AFT random effect model. Here, the random effect is estimated using the hierarchical likelihood (h-likelihood) which is useful for the inference of random effects. The proposed method is demonstrated using simulation studies and a real data set.

임의효과를 이용한 충남지역 소나무림의 바이오매스 모형 개발

표정기 ( Jungkee Pyo ),손영모 ( Yeong Mo Son ) 한국임학회 2017 한국산림과학회지 Vol.106 No.2

본 연구의 목적은 임의효과(random effect)를 이용하여 충남지역 임령-바이오매스 모형을 개발하고 임의효과의 적용성을 평가하는데 있다. 충남지역 소나무림의 임령에 따른 바이오매스 모형 개발을 위해 임분 구조를 고려하여 전국의 중부지방소나무 임분에서 30개소(150그루)를 조사하고 임령과 바이오매스 자료를 수집하였다. 모형 개발에서 중부지방소나무의 임령-바이오매스 관계는 고정효과(fixed effect)이고 지역간 차이를 임의효과로 설정하였다. 임의효과에 따른 모형의 적합도를 검정하기 위해 아카이케의 정보기준(Akaike Information Criterion, AIC)을 참고하고 지역간 차이에 따른 분산-공분산 행렬과 오차항을 추정하였다. 추정된 공분산은 -1.0022, 오차항은 0.6240이고 분산-공분산 행렬을 이용한 임의효과모형의 AIC는 377.7을 나타내어 선행 연구와 이질적인 차이는 없었다. 이러한 결과는 범주형 자료의 임의효과가 모형 개발에 반영된 결과로 판단된다. 본 연구의 결과는 임의효과를 이용하여 일부지역에 국한되어 개발되었던 바이오매스 모형연구에 활용이 가능하다. The purpose of this study was to develop age-biomass model in Chungnam region containing random effect. To develop the biomass model by species and tree component, data for Pinus densiflora in central region is collected to 30 plots (150 trees). The mixed model were used to fixed effect in the age-biomass relation for Pinus densiflora, with random effect representing correlation of survey area were obtained. To verify the evaluation of the model for random effect, the akaike information criterion (abbreviated as, AIC) was used to calculate the variance- covariance matrix, and residual of repeated data. The estimated variance-covariance matrix, and residual were -1.0022, 0.6240, respectively. The model with random effect (AIC=377.2) has low AIC value, comparison with other study relating to random effects. It is for this reason that random effect associated with categorical data were used in the data fitting process, the model can be calibrated to fit the Chungnam region by obtaining measurements. Therefore, the results of this study could be useful method for developing biomass model using random effects by region.

A Random-Effect Prediction Approach for Clustered Binominal Data

조건호 한국자료분석학회 2010 Journal of the Korean Data Analysis Society Vol.12 No.1

Clustered data may have not only a correlation within-cluster but also a heterogeneity between-cluster. Thus, various random-effect models including generalized linear mixed models(GLMMs) have been widely used. However, the analysis of heterogeneity has been less studied. For this the prediction of random effects is very useful. In this paper we show a general framework how to model the clustered binominal data, and to estimate parameters and to predict random effects via SAS PROC NLMIXED. Thus, we illustrate the analyses with a real data set obtained from different 8 clinics for investigating a treatment effect. We demonstrate using this data set how an appropriate model can be selected via well-known model-selection criteria. We also present a plot on predicted random effects for investigating practically a potential heterogeneity over clinics.

주변화 변량효과모형의 조사 및 고찰

전주영,이근백 한국데이터정보과학회 2014 한국데이터정보과학회지 Vol.25 No.6

Longitudinal categorical data commonly occur from medical, health, and social sciences. In these data, the correlation of repeated outcomes is taken into account to explain the effects of covariates exactly. In this paper, we introduce marginalized random effects models that are used for the estimation of the population-averaged effects of covariates. We also review how these models have been developed. Real data analysis is presented using the marginalized random effects. 경시적 범주형자료 (longitudinal categorical data)는 의학, 보건학, 그리고 사회과학에서 많이 발생하는 자료이다. 이러한 자료는 반복측정으로 인한 결과치들의 상관관계를 설명하면서 공변량의 효과를 설명해야 한다. 이 논문에서 모집단에 대한 공변량의 효과를 추정하면서 우도함수에 기초한 모형인 주변화 변량효과모형 (marginalized random effects model)을 소개하고, 그 모형의 어떻게 발전했는지를 고찰한다. 그리고 실제 자료를 이용하여 제시된 모형을 설명한다.

Joint hierarchical generalized linear models with multivariate Gaussian random effects

Molas, M.,Noh, M.,Lee, Y.,Lesaffre, E. North-Holland Pub. Co ; Elsevier Science Ltd 2013 Computational statistics & data analysis Vol.68 No.-

Likelihood based inference for correlated data involves the evaluation of a marginal likelihood integrating out random effects. In general this integral does not have a closed form. Moreover, its numerical evaluation might create difficulties especially when the dimension of random effects is high. H-likelihood inference has been proposed where the explicit evaluation of the integral is avoided. The approach also allows extensions handling e.g. (1) complex design experiments, (2) REML type of inference beyond the class of a linear model and (3) overdispersion modeling. The h-likelihood approach to multivariate generalized linear mixed models is extended. The h-likelihood computational algorithms is blended with a Newton-Raphson procedure for the estimation of the correlation parameters. This allows that components of the joint model are interlinked via correlated Gaussian random effects. Further, correlated random effects are allowed within each component. This approach can serve as a basis for further developments of joint double hierarchical generalized linear models with correlated random effects. The methods are illustrated with a rheumatoid arthritis study dataset, where the correlation between latent trajectories of three endpoints is evaluated.

단순 확산과정들에 대한 확률효과 모형

이은경,이인석,이윤동 한국통계학회 2018 응용통계연구 Vol.31 No.6

Diffusion is a random process used to model financial and physical phenomena. When we construct statistical models for repeatedly observed diffusion processes, the idea of random effects needs to be considered. In this research, we introduce random parameters for an Ornstein-Uhlenbeck diffusion model and geometric Brownian motion diffusion model. In order to apply the maximum likelihood estimation method, we tried to build likelihoods in closed-forms, by assuming appropriate distributions for random effects. We applied the random effect models to data consisting of Dow Jones Industrial Average indices recorded daily over 27 years from 1991 to 2017. 확산은 금융이나 물리적 현상의 모형화에 이용되는 확률과정이다. 반복적으로 관측된 확산과정에 대하여 통계적인 모형을 구축할 때, 확률효과를 고려할 필요가 있다. 이 연구에서는 Ornstein-Uhlenbeck 확산모형과 geometric Brownian motion 확산모형에 대하여 확률효과를 도입한다. 모형모수에 대한 최도우도추정법을 적용하기 위하여, 확률효과에 대한 적절한 분포를 가정하여 닫힌 형태로 우도함수를 얻는 방법을 탐색하였다. 1991년부터 2017년까지 27년간 일일 단위로 기록된 다우존스 산업지수에 대하여 확률효과 모형을 적용하였다.

Statistical prediction of fuel cell catalyst effectiveness: Quasi-random nano-structural analysis of carbon sphere-supported platinum catalysts

Shin, S.,Kim, A.R.,Um, S. Pergamon Press ; Elsevier Science Ltd 2016 International journal of hydrogen energy Vol.41 No.22

<P>In this study, a three-dimensional material network model is developed to visualize the nanoscale structures of carbon sphere-supported platinum (PVC) catalysts and to examine the effective transport paths to optimize the performance of randomly disordered, ternary phase fuel cell catalysts. The catalyst layer domain is modeled using a quasi-random stochastic Monte Carlo-based method that utilizes random number generation processes. Successful interconnections of the three catalyst components are identified, and the catalyst effectiveness is defined to statistically estimate the fraction of the fuel cell catalysts that are utilized. Various fuel cell catalyst compositions are simulated to elucidate the effects of the electron, ion, and mass transport paths on the catalyst effectiveness. The statistical data show that at low ionomer contents, the accessible pore ratio is maximized, enhancing mass transport, and the effective ionomer configuration therefore significantly affects the catalyst effectiveness. In contrast, at high ionomer volume fractions, the ionomers form agglomerate chains that effectively transport ions, whereas the average accessible pore ratios are relatively low. More importantly, this study reveals that the maximum effectiveness depends strongly on the accessible pore ratio and the optimal ionomer volume fraction is inversely proportional to the PVC volume fraction. (C) 2016 Hydrogen Energy Publications LLC. Published by Elsevier Ltd. All rights reserved.</P>

산재근로자 직업재활서비스 지원이 동태적 고용성과에 미치는 영향

김지원 한국행정학회 2018 韓國行政學報 Vol.52 No.2

This study examines the effects of existing occupational rehabilitation policies on the achievement of dynamic employment, which is the upward return to work. It is necessary to classify the upward return-to-work path, which can be regarded as both the quantitative and qualitative aspects of typical employment performance, into two types: 1) narrow dynamic employment performance(new entry into original or other job) and 2) broad dynamic employment performance(new entry into original or other job and retention of original or other job). Using the 4th Workers’ Compensation Insurance Panel Data and applying the random-effects panel logit model, we analyzed the effects of vocational rehabilitation policy variables on achievement of the above two types of employment outcome, while controlling for external characteristics such as individual characteristics of industrial accidents. It was found that the occupational rehabilitation service of industrial accident workers is effective in achieving a narrow dynamic employment performance and that it does not have a significant effect on the achievement of broader dynamic employment outcomes(including employment stability). The main policy suggestions are based on the results of the policy effectiveness tests and the analysis of the differences in policy demand characteristics between new entrants into the job and retention after return to work. 본 연구는 현존 산재근로자 직업재활 정책이 상향적 직업복귀라는 동태적 고용성과 달성에 미치는 영향을 검증한다. 양적, 질적측면의 대표적 고용성과로 볼 수 있는 상향적 직업복귀 경로를 좁은 의미의 동태적 고용성 과(‘원・타직 신규진입’)와 넓은 의미의 동태적 고용성과(‘원・타직 신규진입 또는 고용유지’)의 두 가지로 유형 화한 후 산재근로자 개인 특성 등 외재적 특성을 통제한 상태에서 직업재활 정책 변수가 상기 두 유형의 고용 성과 달성에 미치는 영향을 분석하였다. 제4차 산재보험패널 데이터를 활용하고 성향점수매칭(Propensity Score Matching) 및 확률효과 패널로짓 모형(Random Effect Panel Logit Model)을 적용하여 분석한 결과, 산재근로자 직업재활서비스는 협의의 동태적 고용성과 달성에는 효과적인 반면 고용안정성을 포함하는 넓은 의미의 동태적 고용성과 달성에는 유의미한 영향을 미치지 못하는 것으로 나타났다. 직업재활서비스 정책효과성 실증 분석결과를 토대로 주요 정책제언을 하였다.

Analysis of Multivariate Binary Random Effect Models using Hierarchical Likelihood Approach

김아름,노맹석 한국자료분석학회 2019 Journal of the Korean Data Analysis Society Vol.21 No.4

This paper proposes using hierarchical-likelihood estimation method for binary panel data models featuring state dependence and unmeasured heterogeneity. Hierarchical generalized linear models are an extension of generalized linear models in that they combine generalized linear models with random effects. Hierarchical likelihood based upon hierarchical generalized linear models provides a useful tool for analyzing multivariate data with correlation. These models allow various regression models for mean parameters of response variable as well as dispersion parameters. For a inference, the hierarchical- likelihood approach is suggested for a useful too as model selection and residual plots with real data analysis. In this paper, we suggest hierarchical-likelihood approach estimators for binary random effects-models, and show that this approach outperforms existing marginal likelihood estimators at low computation cost by simulation studies. And then, we add its usefulness by analyzing a real example.