• The Korean Journal of Applied Statistics
• /
• v.33 no.4
• /
• pp.439-449
• /
• 2020

Meta-analysis provides a way of integrating several independent studies of interest. Since small studies with statistically significant results are more likely to be published, publication bias, which is a special case of selection bias, often occurs in meta analysis. Conditional likelihood and weighted estimating equation have been proposed to deal with publication bias, but they require to specify a correct selection probability model. In contrast, the pairwise pseudolikelihood approach can correct publication bias without fully specifying the correct selection probability model, but its performance in meta-analysis was not investigated. In this paper, we perform a numerical study about whether the pairwise pseudolikelihood approach is effective for solving publication bias arising from typical meta-analysis settings.

• The Korean Journal of Applied Statistics
• /
• v.20 no.3
• /
• pp.515-530
• /
• 2007

Stratum-specific likelihood ratio, which is ratio of the sensitivity to 1-the specificity in each stratum of the test, could be biased if the sensitivity and specificity of diagnostic test are affected by verification bias. Therefore, the optimal cutoff point determined by biased stratum-specific likelihood ratios is incorrect. In this study, we derived adjusted stratum-specific likelihood ratios using the adjusted sensitivity and specificity, and obtained the adjusted optimal cutoff point. The influence of the verification bias on the optimal cutoff point was described through the relation between adjusted and unadjusted stratum-specific likelihood ratios.

• Annual Conference on Human and Language Technology
• /
• 2023.10a
• /
• pp.558-562
• /
• 2023

최근 ChatGPT의 등장으로 텍스트 생성 모델에 대한 관심이 높아지면서, 텍스트 생성 태스크의 성능평가를 위한 지표에 대한 연구가 활발히 이뤄지고 있다. 전통적인 단어 빈도수 기반의 성능 지표는 의미적인 유사도를 고려하지 못하기 때문에, 사전학습 언어모델을 활용한 지표인 BERTScore를 주로 활용해왔다. 하지만 이러한 방법은 사전학습 언어모델이 학습한 데이터에 존재하는 편향으로 인해 공정성에 대한 문제가 우려된다. 이에 따라 한국어 사전학습 언어모델의 편향에 대한 분석 연구가 필요한데, 기존의 한국어 사전학습 언어모델의 편향 분석 연구들은 사회에서 생성되는 다양한 속성 별 편향을 고려하지 못했다는 한계가 있다. 또한 서로 다른 언어를 기반으로 하는 사전학습 언어모델들의 속성 별 편향을 비교 분석하는 연구 또한 미비하였다. 이에 따라 본 논문에서는 한국어 사전학습 언어모델의 속성 별 편향을 비교 분석하며, 영어 사전학습 언어모델이 갖고 있는 속성 별 편향과 비교 분석하였고, 비교 가능한 데이터셋을 구축하였다. 더불어 한국어 사전학습 언어모델의 종류 및 크기 별 편향 분석을 통해 적합한 모델을 선택할 수 있도록 가이드를 제시한다.

• Journal of The Korean Association For Science Education
• /
• v.36 no.6
• /
• pp.935-946
• /
• 2016

This study aims to explore cognitive biases relating the core competences of science and instructional strategy in reducing the level of cognitive biases. The literature review method was used to explore cognitive biases and science education experts discussed the relevance of cognitive biases to science education. Twenty nine cognitive biases were categorized into five groups (limiting rational causal inference, limiting diverse information search, limiting self-regulated learning, limiting self-directed decision making, and category-limited thinking). The cognitive biases in limiting rational causal inference group are teleological thinking, availability heuristic, illusory correlation, and clustering illusion. The cognitive biases in limiting diverse information search group are selective perception, experimenter bias, confirmation bias, mere thought effect, attentional bias, belief bias, pragmatic fallacy, functional fixedness, and framing effect. The cognitive biases in limiting self-regulated learning group are overconfidence bias, better-than-average bias, planning fallacy, fundamental attribution error, Dunning-Kruger effect, hindsight bias, and blind-spot bias. The cognitive biases in limiting self-directed decision-making group are acquiescence effect, bandwagon effect, group-think, appeal to authority bias, and information bias. Lastly, the cognitive biases in category-limited thinking group are psychological essentialism, stereotyping, anthropomorphism, and outgroup homogeneity bias. The instructional strategy to reduce the level of cognitive biases is disused based on the psychological characters of cognitive biases reviewed in this study and related science education methods.

• Annual Conference on Human and Language Technology
• /
• 2023.10a
• /
• pp.533-538
• /
• 2023

문서에서 저자의 의도와 주제, 그 안에 포함된 감성을 분석하는 것은 자연어 연구의 핵심적인 주제이다. 이와 유사하게 특정 글에 포함된 정치적 문화적 편향을 분석하는 것 역시 매우 의미 있는 연구주제이다. 우리는 최근 발생한 한 사건에 대하여 여러 신문사와 해당 신문사에서 생산한 기사를 중심으로 해당 글의 정치적 편향을 정량화 하는 방법을 제시한다. 그 방법은 선택된 주제어들의 문장 공간에서의 거리를 중심으로 그래프를 생성하고, 생성된 그래프의 기계학습을 통하여 편향과 특징을 분석하였다. 그리고 그 그래프들의 시간적 변화를 추적하여 특정 신문사에서 특정 사건에 대한 입장이 시간적으로 어떻게 변화하였는지를 동적으로 보여주는 그래프 애니메이션 시스템을 개발하였다. 실험을 위하여 최근 이슈에 대하여 12개의 신문사에서 약 2000여 개의 기사를 수집하였다. 그 결과, 약 82%의 정확도로 일반적으로 알려진 정치적 편향을 예측할 수 있었다. 또한, 학습 데이터에 쓰이지 않은 신문기사를 활용하여도 같은 정도의 정확도를 보임을 알 수 있었다. 우리는 이를 통하여 신문기사에서의 정치적 편향은 작성자나 신문사의 특성이 아니라 주제어들의 문장 공간에서의 거리 관계로 특성화할 수 있음을 보였다. 할 수 있다.

• The Korean Journal of Applied Statistics
• /
• v.17 no.3
• /
• pp.459-473
• /
• 2004

It has well known that an exhaustive search algorithm suggested by Breiman et. a1.(1984) has a trend to select the variable having relatively many possible splits as an splitting rule. We propose an algorithm to overcome this variable selection bias problem and then construct unbiased regression trees based on the algorithm. The proposed algorithm runs two steps of selecting a split variable and determining a split rule for binary split based on the split variable. Simulation studies were performed to compare the proposed algorithm with Breiman et a1.(1984)'s CART(Classification and Regression Tree) in terms of degree of variable selection bias, variable selection power, and MSE(Mean Squared Error). Also, we illustrate the proposed algorithm with real data sets.

• Proceedings of the Korean Information Science Society Conference
• /
• 2012.06c
• /
• pp.314-316
• /
• 2012

대용량 지형 데이터를 실시간에 렌더링 하기 위해 여러 가지 연속상세단계 기법들이 연구되었다. 하지만 이러한 방법을 적용해도 지형 데이터가 하드웨어에서 처리할 수 있는 크기보다 클 경우 과도한 간략화로 인한 기하오차가 발생하거나 프레임률이 저하된다. 또한 기존 연속상세단계 기법을 수행하기 위해 만들어진 자료구조들 또한 지형 데이터의 크기에 비례하여 커지므로 메모리와 전처리 시간이 많이 소요된다. 본 논문에서는 적은 개수의 정점으로 효과적인 지형 렌더링이 가능한 편향맵을 다해상도로 확장하여 별도의 자료구조가 따로 필요 없는 간단한 연속상세단계 기법을 제안한다. 이 방법은 적은 메모리 용량으로 높은 정확도의 지형을 실시간에 렌더링 할 수 있다. 연속상세단계 선택은 보다 빠른 처리를 위해 GPU에서 패치 단위의 테셀레이션을 통해서 단일 패스로 수행된다. 상세단계가 선택으로 세분화 된 지형의 각 정점들은 화면 공간상의 오차를 참조하여 각각의 상세단계를 선택한 후 해당되는 편향맵에 저장된 이동벡터만큼 이동하여 최종 지형 메쉬를 생성한다. 제안한 방법은 전처리 단계를 포함한 모든 처리가 GPU에서 수행되므로 속도가 빠르고 적은 정점으로 보다 정확한 지형을 렌더링 할 수 있다.

• Journal of Digital Convergence
• /
• v.19 no.11
• /
• pp.135-144
• /
• 2021

This study investigates the underlying mechanism of YOLO in millennial consumer, especially the influence of perceived economic recession on the present-biased preference. In addition, it was attempted to expand the implicit theory by proposing the individual's entity belief as a mediator for the effect of perceived economic situation on consumers' present-biased seeking behavior. In three experimental studies, undergraduate students who both highly primed and measured negative economic situation showed more favorable attitudes towards present-biased persuasive message and related products. The results of this research provides practical implication for marketers especially in the current situation experiencing economic slowdown due to low economic growth and COVID 19.

• Convergence Security Journal
• /
• v.21 no.1
• /
• pp.201-208
• /
• 2021

The purpose of this study is to investigate reunification perception among groups with different cognitive bias(anchoring bias) through empirical studies. This study assumes that cognitive bias occurs according to the content of N. Korea provided in school education, and that the cognitive bias formed in youth creates a difference in perception of N. Korea and reunification. For empirical study, a survey of reunification awareness conducted by the Institute for Peace an Unification Studies-Seoul National University was used, and the two groups are divided into A-group that chose "anti-communism or security" and B-group that did "reunification interests, multi-culturalism and understanding North Korea" according to N. Korea content which was encountered in school education during adolescence. A-group recognized N. Korea as a hostile target even after becoming an adult, while B-group regarded N. Korea as a one of dialogue and cooperation. This analysis illustrates the existence of cognitive bias between the two groups because of N. Korea content which was given in school education. A-group had a lower percentage than B-group in terms of the benefits of reunification to S. Korea and the need for more dialogue and compromise with N. Korea. However, there was no difference between two groups in terms of acceptance of N. Korean defectors. This research result could be a reference to the direction of school education policies relating to reunification in that early school education affects reunification awareness and perception related with N. Korea even in adulthood.

• Journal of Digital Convergence
• /
• v.19 no.1
• /
• pp.355-360
• /
• 2021

In this study, the factors of confirmation bias that may occur in the initial user experience stage were analyzed using a honeycomb model by deriving user experience factors for each factor. In the initial user experience stage, confirmation bias occurs in the impression stage. At the processing stage of memory, sensory memory, working memory, and long-term memory, which stores and retrieves selective memory, were closely related. Confirmation bias was classified into visibility, correlation, memory, clarity, and universality in the usability part, and satisfaction, joy, and dissatisfaction were derived as emotional factors. As a result of the analysis with the honeycomb model, visuality, clarity, universality in the usability factor, and joy in the emotional factor had little effect on the confirmation bias, and satisfaction and dissatisfaction were analyzed as the main factors of the confirmation bias in the correlation, memory and emotional factors. This study is meaningful in that it can be usefully used as a reference material for companies that customize design patterns for the factor of confirmation bias.