• 응용통계연구
• /
• 제34권1호
• /
• pp.1-8
• /
• 2021

사회조사에서 추상적 개념은 타당성과 신뢰성이 보장된 측정도구를 사용하여야 한다. 그와 같은 측정도구에 의해 도출된 관측점수는 타당한 관측점수와 편향된 관측점수와 오차로 나눌 수 있으며 편향된 값의 유무가 타당성과 연관되어 있으며 오차값의 유무가 신뢰성과 연관되어 있다. 측정도구가 타당도 및 신뢰도를 만족하는 지를 보기 위한 기법들이 많이 존재한다. 예를 들면 요인분석을 통한 구성타당도, 크론바흐 알파 값에 의한 내적일치도 등을 들 수 있다. 본 연구에서 크론바흐알파값의 계산은 표본을 통해서 도출되는 데 복잡한 표본설계와 무응답이 발생했을 때 크론바흐알파값의 추정법에 대해 살펴본다. 제안된 기법에 대한 모의실험으로 다변량정규분포를 사용하여 기존의 여러 다른 크론바흐알파값의 추정기법과 비교분석한다.

• 지구물리
• /
• 제5권2호
• /
• pp.131-142
• /
• 2002

일반적으로 지자기 전달함수는 관측치와 예측치의 차이를 최소화하는 관점에서 해가 추정된다. 오차의 구조가 가우스 분포를 따르면 최소자승 추정이 최적의 추정이지만, 그렇지 않은 경우 전달 함수 추정을 심각하게 왜곡시킬 수 있으므로 오차 구조에 대한 정보가 요구된다. 본 연구에서는 Q-Q plot을 이용한 오차 구조으 검증을 통하여 실제 오차 구조에 대한 정보를 획득하였고 가우스 분포 가정을 벗어나는 오차 구조에 대해 외치(outlier)에 의한 영향을 최소로 하며 해를 추정하는 로버스트 추정(regression M-estimate)을 적용하였다. 오차가 가우스 분포를 따르는 경우, 최소자승 추정과 로버스트 추정은 유사한 결과를 나타내나, 오차가 가우스 분포를 벗어나는 경우 로버스트 추정이 최소자승 추정보다 부드러운 결과를 나타냄을 확인하였다.

• 한국정보전자통신기술학회논문지
• /
• 제12권4호
• /
• pp.327-335
• /
• 2019

본 연구는 5년간의 세탁기 및 화장실 용수의 실측 자료를 기반으로 표본가구의 가구 및 주택, 월 특성들을 나타내는 변수들을 조사하여 위의 두 가지 용수에 대한 수요예측모형을 개발하는 것이다. 그러나 반응변수인 세탁기 및 화장실 용수의 분포를 확인한 결과 양의 왜도, 즉, 왼쪽으로 치우친 형태로 정규분포를 따르지 않기 때문에 다중회귀모형 적용 시 추정치가 편의되는 문제가 있다. 따라서 이에 대한 대안으로 세탁기 및 화장실 용수 수요 예측모형을 와이블 및 대수정규회귀모형으로 가정하고, 3가지 모형을 적용하여 최적모형을 제시하고 이에 대한 해석을 제시하였다. 그 결과 와이블 분포를 가정한 회귀모형이 가장 적합하다고 나타났으며, 이에 대한 각 용수별 예측모형을 제시하고 해석하였다. 그 결과 두 용수 모두 공통적으로 실거주인원이 용수사용에 가장 큰 영향을 주는 인자로 나타났다. 분석결과를 토대로 세탁기 및 화장실 용수의 수요를 예측시 실거주인원에 대한 고려가 필요하다는 점을 시사하며, 예측모형을 통한 관리부서에서 장기적으로 물 수요관리에 대한 정책 수립, 수도 관련 시설 규격 및 기자재 결정 등의 기초자료로 활용 할 수 있을 것이라 판단된다. 추후 연구에서는 1인 가구 등과 같은 실태를 반영한 독립변수들을 고려한 실측연구 등이 필요할 것이다.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제17권12호
• /
• pp.3383-3397
• /
• 2023

Scene graphs serve as semantic abstractions of images and play a crucial role in enhancing visual comprehension and reasoning. However, the performance of Scene Graph Generation is often compromised when working with biased data in real-world situations. While many existing systems focus on a single stage of learning for both feature extraction and classification, some employ Class-Balancing strategies, such as Re-weighting, Data Resampling, and Transfer Learning from head to tail. In this paper, we propose a novel approach that decouples the feature extraction and classification phases of the scene graph generation process. For feature extraction, we leverage a transformer-based architecture and design an adaptive calibration function specifically for predicate classification. This function enables us to dynamically adjust the classification scores for each predicate category. Additionally, we introduce a Distribution Alignment technique that effectively balances the class distribution after the feature extraction phase reaches a stable state, thereby facilitating the retraining of the classification head. Importantly, our Distribution Alignment strategy is model-independent and does not require additional supervision, making it applicable to a wide range of SGG models. Using the scene graph diagnostic toolkit on Visual Genome and several popular models, we achieved significant improvements over the previous state-of-the-art methods with our model. Compared to the TDE model, our model improved mR@100 by 70.5% for PredCls, by 84.0% for SGCls, and by 97.6% for SGDet tasks.

• 천문학회보
• /
• 제42권1호
• /
• pp.29.4-30
• /
• 2017

Korea Astronomy and Space Science Institute We present a weak-lensing study of the galaxy cluster SPT-CL J2106-5844 at z=1.132 discovered in the South Pole Telescope Sunyaev-Zel'dovich (SPT-SZ) survey. The cluster is claimed to be the most massive system at z > 1 in the SPT-SZ survey. The inferred mass ($M_{200c}=(1.27{\pm}0.21){\times}10^{15}M_{sun}$) is somewhat unusual at such a high redshift given the current ΛCDM prediction. The mass estimates, however, may be biased because the hydrostatic assumption may not hold when the universe was about 40% of the current age. In this work, we reconstruct the dark matter distribution and measure the mass of this interesting cluster using weak-lensing analysis based on the images from the Advanced Camera for Surveys and Wide Field Camera 3 on-board the Hubble Space Telescope. We find that the mass distribution of the cluster is unimodal with no significant substructures. The centroid of the dark matter agrees with both galaxy luminosity and number density distributions, as well as the hot gas centroid. We confirm that the cluster is indeed extremely massive ($M_{200c}=(1.81{\pm}0.47){\times}10^{15}M_{sun}$) supporting the previous non-lensing measurements. We also discuss the rarity of the cluster in the ΛCDM cosmology, comparing with the expected abundance of similarly massive clusters.

• 유통과학연구
• /
• 제15권11호
• /
• pp.49-57
• /
• 2017

Purpose - In Korea, general super markets have a great impact on the market performance of traditional markets. We propose a modified two stage DEA model for evaluating the performance of traditional markets in Incheon, Korea by identifying the influence of external environmental factors including the presence of general super markets as non-discretionary variables in DEA. Research design, data, and methodology - After obtaining bias-corrected estimates of original DEA efficiency scores using the input and output data of 49 traditional markets, we regress them on several external environmental factors by bootstrap-truncated regression. Results - We obtain bias-corrected efficiency scores from the original DEA efficiency scores by bootstrap and among the five environmental factors, the residential population and the presence of general super markets or SSMs can be considered as the driving forces influencing bias-corrected efficiency scores, positively and negatively, respectively. Conclusions - When DEA efficiency scores tend to be overestimated, we need to use a biased-corrected efficiency score by bootstrap. It is important to note that the efficiency of traditional markets can be largely influenced by external environmental factors such as the presence of general super markets or SSMs that traditional markets can not control. Therefore, it is desirable to consider such environmental factors appropriately for a reasonable performance evaluation.

• 천문학회보
• /
• 제44권1호
• /
• pp.85.1-85.1
• /
• 2019

Accurate estimation of the masses, the ages, and the chemical abundances of host stars is crucial to understand physical characteristics of exo-planetary systems. In this study, we investigate physical dimensions of 94 planet-hosting stars based on spectroscopic observation and stellar evolutionary computation, From the high resolution echelle spectroscopy of the BOES observation, we have analysed metallicities and alpha-element enhancements of host stars. By combining recent spectro-photometric observations, stellar parameters are calibrated within the frame work of the standard stellar theory. In general, the minimum chi-square estimation can be strongly biased in cases that stellar properties rapidly changes after the terminal age main-sequence. Instead, we adopt a Bayesian statistics considering a priori distribution of stellar parameters during the rapid evolutionary phases. we determine a reliable set of stellar parameters between theoretical model grids. To overcome this statistical bias, (1) we adopt a Bayesian statistics considering a priori distribution of stellar parameters during the rapid evolutionary phases and (2) we construct the fine model grid that covers mass range ($0.2{\sim}3.0M_{\odot}$) with the mass step ${\Delta}M=0.01M_{\odot}$, metallicities Z = 0.0001 ~ 0.04, and the helium and the alpha-element enhancement. In this presentation, we introduce our calibration scheme for several hosting stars.

• 유통과학연구
• /
• 제20권12호
• /
• pp.117-123
• /
• 2022

Purpose: This study investigates the relationship between earnings forecasts estimated from a cross-sectional earnings forecast model and firm characteristics such as firm size, sales volatility, and earnings volatility. Research design, data and methodology: The association between earnings forecasts and the aforementioned firm characteristics is examined using 214 firm-year observations with analyst following and 848 firm-year observations without analyst following for the period of 2011-2019. I estimate future earnings using a cross-sectional earnings forecast model, and then compare these model-based earnings forecasts with analysts' earnings forecasts in terms of forecast bias and forecast accuracy. The earnings forecast bias and accuracy are regressed on firm size, sales volatility, and earnings volatility. Results: For a sample with analyst following, I find that the model-based earnings forecasts are more accurate as the firm size is larger, whereas the analysts' earnings forecasts are less biased and more accurate as the firm size is larger. However, for a sample without analyst following, I find that the model-based earnings forecasts are more pessimistic and less accurate as firms' past earnings are more volatile. Conclusions: Although model-based earnings forecasts are useful for evaluating firms without analyst following, their accuracy depends on the firms' earnings volatility.

• Steel and Composite Structures
• /
• 제49권4호
• /
• pp.431-440
• /
• 2023

The functionally graded (FG) porous plates are usually characterized by the non-symmetric elastic modulus distribution through the thickness so that the plate neutral surface does not coincide with the mid-surface. Nevertheless, the conventional analysis models were mostly based on the plate mid-surface so that the accuracy of resulting numerical results is questionable. In this context, this paper presents the neutral surface-based static and free vibration analysis of FG porous plates and investigates the differences between the mid- and neutral surface-based analysis models. The neutral surface-based numerical method is formulated using the (3,3,2) hierarchical model and approximated by the last introduced natural element method (NEM). The volume fractions of metal and ceramic are expressed by the power-law function and the cosine-type porosity distributions are considered. The proposed numerical method is demonstrated through the benchmark experiment, and the differences between two analysis models are parametrically investigated with respect to the thickness-wise material and porosity distributions. It is found from the numerical results that the difference cannot be negligible when the material and porosity distributions are remarkably biased in the thickness direction.

• 한국IT서비스학회지
• /
• 제12권3호
• /
• pp.127-149
• /
• 2013

Is the social web environment in which production, distribution and consumption of information occurs from users an environment where manifestation of collective intelligence is easily made? Or is the social web environment a condition that incites people to depend on the groupthink due to biased information? It is important to conduct empirical studies on the possibility of social media as a tool of collective intelligence under the situation where conflicting opinions prevail. However, most of the existing studies related to this were limited to an exploratory research rather than an empirical research. In this regard, this study attempted to examine if the social media can perform a part as a platform of the collective intelligence empirically. Based on the experimental results, it can be safely said that the communication methods of social media showed its usefulness in both 'intellectual capacity of the group' and 'problem-solving skill of the group.'