Earticle

기존의 고정모수를 이용한 가산모형은 관측되지 않은 이질성을 고려할 수 없으며, 계수값의 표준오차가 과소추정됨에 따라 과도한 t-값이 도출되어 모형의 신뢰성이 감소하게 된다. 또한 무신호교차로의 경우 자료 수집의 어려움과 정확한 분석과정의 통계적 한계로 신호교차로에 비해 연구가 미비한 실정이다. 본 연구에서는 Random Parameter를 이용한 가산모형을 구축하여 교통사고에 영향을 미치는 요인들을 분석하고자 하였으며, 지방부 무신호교차로를 대상으로 하여 기존 연구들과의 차별성을 두고자 하였다. 분석 결과, 총 7개의 변수가 유의한 변수로 나타났고, 이 중 2개의 변수(횡단보도 유무, 제한 속도)가 확률적 변수로 나타났다.

Previous count models using fixed parameter can not consider the unobserved heterogeneity, as the standard error of the count value is underestimated, excessive t-values are derived thereby reducing the reliability of the model. Also, the study of unsignalized intersections are inadequate because of the difficulty of collecting data and statistical limits for accurate analytical processes compared to the signalized intersections. The purpose of this study is to analyze the factors affecting traffic accidents by constructing the count model using random parameters, and it aimed to distinguish between existing studies based on the rural unsignalized intersections. As a result of the analysis, 7 variables were presented as significant variables, and 2 variables(presence of crosswalk, speed limit) were presented as random parameter.

본 연구는 비시장재인 자연환경자원의 방문수요함수를 화폐소득 뿐만 아니라 시 간소득을 포함하는 두 예산제약 하에서 소비자 효용극대화 선택과정으로 전개함 으로써 여가시간의 기회비용(두 라그랑지언 승수의 비율)을 내생적으로 도출하 고, 개인들의 선호와 관련된 인구통계학적 변수들의 함수로 나타내었다. 새만금 방조제의 완성으로 영향을 받을 것으로 예상되는 변산반도 국립공원 방문자들을 대상으로 실시된 방문지점 설문조사 자료를 사용하여, 방문수요함수와 여가시간 의 가치함수를 결합추정하였다. 실증분석 결과 방문수요함수와 여가시간 가치함 수에 포함된 대부분의 변수들이 통계적으로 유의하였다. 국립공원 방문수요행동 에 함축된 여가시간에 대한 가치는 취업자들의 경우 시장임금율의 55%정도였고, 비취업자들도 여가시간에 대해 시장임금율의 20%정도의 잠재적 가치를 부여하고 있는 것으로 나타났다. 또한 방문자들이 변산반도 국립공원 1회 방문을 통하여 느끼는 소비자 잉여는 여가시간의 가치의 포함여부에 따라 31,500원과 56,700원 으로 통계적으로 유의하게 차이가 났다.

This paper is to estimate values of leisure time revealed by recreational demand choices for natural resources such as national parks. Marginal value of leisure time can be endogenously derived as a ratio of two Lagrange multipliers in the process of solving a utility-consistent model of choice subject to two constraints(both money and time budgets). Both recreational demand function and value of leisure time, which was specified as a function of individual characteristics, were jointly estimated in an application to Byunsan Peninsula National Park near the Saemankum area. The value of leisure time was measured to be about 55% of market wage for workers and about 20% for non-workers. Moreover, consumer surplus per visit was higher when value of leisure time was incorporated explicitly. The empirical approach can be applied with only minor additions to conventional survey techniques for nonmarket valuation.

백혈구 수치는 체내 염증 수치 및 패혈성 쇼크를 알 수 있는 중요한 인체정보이다. 그러나 백혈구 수치는 혈액검사가 필요하며, 검사결과를 확인하기 위해서는 2시간의 의료대기 시간이 필요하다. 따라서 응급한 환자들은 혈 액검사 결과를 기다리는 동안에 의료적 대응을 받기 어려울 수 있다. 이런 문제를 해결하기 위해 현장에서는 생체신 호 및 정보를 바탕으로 의료적 대응을 수행하지만, 정량적으로 제공되는 백혈구 수치에 따른 대응과는 차이가 있다. 따라서 본 연구에서는 인공지능의 GAN(Generative Adversarial Networks)을 기반으로 환자의 생체신호 및 정보를 입력받아, 백혈구 수치를 예측하는 인공지능 모델에 관한 연구를 수행했다. 모델의 검증은 원본 데이터와 결측 데이 터를 혼합하여, 인공지능 모델에 입력했을 때의 성능으로 구성했다. 검증 결과로는 백혈구 예측모델을 통해 생성된 백혈구 수치를 포함한 혼합 데이터군이 원본 데이터군보다 패혈증 환자에 대한 검출 성능이 뛰어난 것으로 나타났다.

Leukocyte levels are important human information to know inflammatory levels in the body and septic shock. However, white blood cell levels require a blood test, and two hours of medical waiting time is required to confirm the test results. Therefore, emergency patients may find it difficult to receive medical responses while waiting for blood test results. In order to solve this problem, medical responses are performed based on bio-signals and information in the field, but there is a difference from responses according to quantitatively provided white blood cell levels. Therefore, in this study, we conducted a study on an artificial intelligence model that predicts white blood cell levels by receiving patient bio-signals and information based on Generative Adversarial Networks (GAN) of artificial intelligence. The verification of the model consisted of the performance when input into the artificial intelligence model by mixing the original data and missing data. The verification results showed that the mixed data group including white blood cell levels generated through the white blood cell prediction model had better detection performance for sepsis patients than the original data group.

It has been known generally that investment in the extending ability to use the IT applications have further enhanced the productivity of effects of IT on firms by reducing costs, increasing returns, and increasing the speed of operations, etc. Notwithstanding this fact, it was very complex and difficult to evaluate concretely the effect of informatization of firm. SMEs(Small- & Medium-sized Enterprises) in particular. In this study, I point out the weakness of SMEs and analyze the effects of informatization through the count data model. For this analysis, I separate the effects into two part, such as organizational effect and personal effect. It comes to conclusion that organizational effect is larger than personal effect and the ability to practical use of IT systems is most efficient item related with informatization level. Since it will be important to cencentrate on raising this ability for heightening the competitiveness of SMEs.

개인교통수단의 선호로 인한 자가용 승용차의 급증은 서울시의 교통혼잡을 가중시키는 주요한 요인이 되고 있다. 이러한 서울시의 교통혼잡을 완화하기 위해서는 대중교통 중심의 교통체계가 구축되어야 하며 승용차 이용자를 대중교통수단으로 유인할 수 있는 대중교통 활성화정책이 필요하다. 이러한 인식하에 버스를 이용하는 통근 및 통학목적 통행자의 버스이용횟수에 대한 개별행태모형을 통하여 버스 이용에 영향을 미치는 요인을 파악함으로써 승용차 이용자를 대중교통수단으로 유인할 수 있는 정책적인 시사점을 도출하고자 하였다. 본 연구의 목적은 일주일간 버스이용횟수 추정에 적합한 가산자료모형의 적용이다. 국내에서는 가산자료모형을 이용한 연구가 많지 않은 실정이며, 또한 모형의 설정시 과산포(overdispersion)에 대한 검정을 통하여 자료에 적합한 모형을 설정하는 것이 중요함에도 불구하고 적절한 검정없이 일반적으로 사용되고 있는 포와송 회귀모형을 주로 사용하여 왔다. 그러나 본 연구에서는 가산자료모형을 선정하기 전에 과산포에 대한 통계적인 검정을 시행한 결과 음이항 회귀모형이 본 연구의 자료에 적합한 것으로 판정되었으며, 모형설정의 중요성을 살펴보기 위하여 음이항 회귀모형을 이용하여 추정한 결과와 포와송 회귀모형을 이용하여 추정한 결과를 비교하여 보았다.

The rapid increase of Passenger cars which is caused by the discomfort of Public transit and the Preference of automobiles is the major factor of increasing traffic congestions in Seoul With the point that leading the automobilists to the Public transit can be the most important Policy to ease these traffic congestions, this study focuses on the behavioral aspects of company employees and university students and investigates factors influencing bus ridership. To be brief, by estimating bus ridership through count models, this study investigates factors which influence bus ridership and elicits Political suggestions which lead automobilists to Public transit. The Purpose in this study is the application of appropriate count data model. The count data models have been widely applied to the economic area from the middle of the 1980s and to transportation aspect mainly in the foreign countries from the latter half of the 1980s. Even though a few studies in this country employed count data model to count data. all of them were Poisson regression models without suitable tests for the importance of the model specification. In the end, as the result of statistical test, negative binomial regression model which is suitable for overdispersed data was found to be appropriate for the data of weekly bus ridership. To emphasize the importance of model specification, both of poisson regression model and negative binomial regression model were estimated and the results were compared.

For count responses, the situation of excess zeros(disproportionately more zeros than what standard models allow) often occurs in management and sociological applications. The zero-inflated Poisson regression model has been used for count data with excess zeros. An alternative approach is the hurdle model, in which the second part is governed by a truncated Poisson or negative binomial distribution. We apply the models to draw the relation between the frequency of moving mobile phone companies and customers' social attributes such as age, income and so on. We compare the models to the Poisson model in terms of log-likelihood.

The methodology to identify the traffic count stations that have anomalous counts was developed in this paper. Our proposed approach is to develop a spatial model that would separate out systematic and spatial variability in counts so that residual uncertainty may be identified. A principal component analysis was applied by using dimension reduction dealing with census data with many attributes. This proposed method would significantly improve the process of validating traffic counts by increasing the accuracy of reported counts and by reducing the time delay between data collection and reporting.

Purpose The study investigated whether a replacement of neutrophil count and platelet count by neutrophil-to-lymphocyte ratio (NLR) and platelet-to-lymphocyte ratio (PLR) within the International Metastatic Renal Cell Carcinoma Database Consortium (IMDC) model would improve its prognostic accuracy. Materials and Methods This retrospective analysis included consecutive patients with metastatic renal cell carcinoma treated with first-line tyrosine kinase inhibitors. The IMDC and modified-IMDC models were compared using: concordance index (CI), bias-corrected concordance index (BCCI), calibration plots, the Grønnesby and Borgan test, Bayesian Information Criterion (BIC), generalized R2, Integrated Discrimination Improvement (IDI), and continuous Net Reclassification Index (cNRI) for individual risk factors and the three risk groups. Results Three hundred and twenty-one patients were eligible for analyses. The modified-IMDC model with NLR value of 3.6 and PLR value of 157 was selected for comparison with the IMDC model. Both models were well calibrated. All other measures favoured the modified-IMDC model over the IMDC model (CI, 0.706 vs. 0.677; BCCI, 0.699 vs. 0.671; BIC, 2,176.2 vs. 2,190.7; generalized R2, 0.238 vs. 0.202; IDI, 0.044; cNRI, 0.279 for individual risk factors; and CI, 0.669 vs. 0.641; BCCI, 0.669 vs. 0.641; BIC, 2,183.2 vs. 2,198.1; generalized R2, 0.163 vs. 0.123; IDI, 0.045; cNRI, 0.165 for the three risk groups). Conclusion Incorporation of NLR and PLR in place of neutrophil count and platelet count improved prognostic accuracy of the IMDC model. These findings require external validation before introducing into clinical practice.

This paper considers a specification test of conditional Poisson regression model for time series count data. Although conditional models for count data have received attention and proposed in several ways, few studies focused on checking its adequacy. Motivated by the test of martingale difference assumption, a specification test via Ljung-Box statistic is proposed in the conditional model of the time series count data. In order to illustrate the performance of Ljung- Box test, simulation results will be provided.

In many applications, we frequently encounter correlated multiple outcomes measured on the same subject. Joint modeling of such multiple outcomes can improve efficiency of inference compared to independent modeling. For instance, in developmental toxicity studies, fetal weight and number of malformed pups are measured on the pregnant dams exposed to different levels of a toxic substance, in which the association between such outcomes should be taken into account in the model. The number of malformations may possibly have many zeros, which should be analyzed via zero-inflated count models. Motivated by applications in developmental toxicity studies, we propose a Bayesian joint modeling framework for continuous and count outcomes with excess zeros. In our model, zero-inflated Poisson (ZIP) regression model would be used to describe count data, and a subject-specific random effects would account for the correlation across the two outcomes. We implement a Bayesian approach using MCMC procedure with data augmentation method and adaptive rejection sampling. We apply our proposed model to dose-response analysis in a developmental toxicity study to estimate the benchmark dose in a risk assessment.

A Poisson error model as a generalized linear mixed model (GLMM) has been suggested for genetic analysis of counted observations. One of the assumptions in this model is the normality for random effects. Since this assumption is not always appropriate, a more flexible model is needed. For count traits, a Poisson hierarchical generalized linear model (HGLM) that does not require the normality for random effects was proposed. In this paper, a Poisson-Gamma HGLM was examined along with corresponding analytical methods. While a difficulty arises with Poisson GLMM in making inferences to the expected values of observations, it can be avoided with the Poisson-Gamma HGLM. A numerical example with simulated embryo yield data is presented.

This study investigated the e-commerce platform food delivery O2O service of restaurant companies. Recognizing the need for ways to maximize the efficiency of marketing activities, a demand model for the O2O platform consumption market was established. The purpose of the study is to understand the behavior of consumers using the O2O platform service of restaurant companies through demand determinants estimation, and to understand the determinants that affect the demand of consumers in the O2O platform service market. The truncated Poisson model (TP) was used as the demand model, and the analysis was conducted by applying the truncated negative binomial model (TNB) as an alternative model for fear of overdispersion. As a result of the analysis, both the truncated Poisson model and the truncated negative binomial model were found to be statistically significant. Since the dependent variable analyzed in this study has a statistically significant problem of transient variance, the TNB model was applied in this study and the explanatory variable for the demand model was interpreted accordingly. Accordingly, the TNB model considering overdispersion showed a more improved fit of the model than the TP model, and significant variables were identified in the O2O platform demand determinants, focusing on the TNB model. As a result of the estimation of the demand model, the variables that have a positive (+) effect on the demand for the O2O platform of restaurant companies were unmarried, single-person households, and snack menus were preferred, and convenience and information factors were found in the O2O platform service attribute factors. As variables influencing negative (-), the form menu and the O2O platform service attribute factors were investigated as economic factors. This study will complement the objectivity of the research by conducting an empirical study through a survey that diversifies and segments the target audience of O2O platforms for food service companies, and we hope to conduct a study that separates food service e-commerce platforms.

본 연구의 목적은 카지노방문객의 수요모형을 통해 결정요인들을 확인해내는데 있다. 이를 위하여 본 연구에서는 크게 종속변수인 방문횟수가 가산자료일 때 적용될 수 있는 절단된 포아송(TPOIS)모형과 절단된 음이항(TNB)모형을 이용하였다. 특히, 선행연구들이 종속변수를 단일기간으로만 측정하였다는 점에 착안하여 본 연구에서는 측정기간을 지난 6개월간 방문횟수, 1년간 방문횟수, 3년간 방문횟수로 세분화하였다. 또한, 설명변수로는 기존 카지노방문객 수요모형 연구에서 사용되었던 인구통계학적 특성 외에도 사회심리학 분야에서 인간행동을 이해하기 위해 연구되어온 계획행동이론의 변수들(태도, 주관적 규범, 지각된 행동통제감)을 모형에 추가하였다. 이러한 변수들을 통해 보다 다양한 측면에서 카지노방문객들의 방문수요 특성을 이해하고자 하였다. 모형 추정결과, 종속변수를 ‘지난 3년간 방문횟수’로 설정한 TNB모형이 가장 적합한 것으로 나타났다. 또한, 모형을 통해 추정된 카지노방문객 수요결정요인은 다음과 같다. 카지노 방문객이 남성일수록, 소득수준이 높을수록, 주요 방문목적이 카지노 방문일수록, 카지노에 대한 사전 지식정도가 높을수록, 카지노 방문수요에 정(+)의 영향을, 주관적 규범은 카지노 방문수요에 부(-)의 영향을 미치는 것으로 나타났다.

This study aims to identify determinants of demand for casino visitors. To this end Truncated Poisson(TPOIS) Model and Truncated Negative Binomial(TNB) Model were used because the dependent variable of this study was count data truncated zero. In order to understand the characteristics of casino visitors constructs based on theory of planned behavior(TPB) were also considered as independent variables in the model along with demographic and situational-specific variables. Furthermore, time periods of the dependent variable (frequency of visiting a casino) were specified with three categories(during the last 6 months, one year, and three years) because previous studies only measured the dependent variable using a single period. The results of this study reveal that the TNB Model with the dependent variable of “the frequency of visiting a casino during the last three years” was preferred to other two models with 6 months and 1 year visit frequency. The results of this study also indicate that gender, income, main purpose of visiting casino, and prior knowledge had significantly positive effect on demand for casino visitors, whereas subjective norm had a significantly negative effect on demand for casino visitors.

This study investigates the determinants of output of a manufacturing firm's innovative activity (the number of patent applications) and spillover effects using a count data model. This paper attempted a negative binomial distribution In order to take into account unobserved heterogeneity. The results of our study suggested that Firm size, R&D intensity, technical network activity, and online business performance have significantly positive effects in the Korean manufacturing industry. Moreover, there are significantly positive spillover effects in the same industry sector.

연구목적: 본 연구는 중저가 와인을 대상으로 와인 소비자의 구매행동을 분석하는 것에 목적을 두었다. 연구설계 및 방법론: 연구목적을 달성하기 위해 종속변수인 구매횟수(지난 6개월간 와인 구매 횟수, 지난 1년간 와인 구매 횟수)의 자료특성에 따라 적절한 모형으로 절단된 포아송(TPOIS)모형과 절단된 음이항(TNB)모형을 설정하고, 구매횟수에 영향을 미치는 영향요인들을 도출하였다. 분석 및 연구 결과: 모형 추정결과, 총 4개 모형 간 추정된 계수값은 아주 큰 차이는 없지만, 전반적으로 종속변수를 ‘지난 6개월간 와인 구매횟수’로 선정한 TPOIS모형, TNB모형의 설명변수가 더 유의한 것으로 나타났으며 이론적으로 더 타당한 모형인 TNB모형을 중심으로 와인 소비자의 구매행동을 분석하였다. 구체적인 분석결과는 다음과 같다. 첫째, 인구통계적 특성 중에서는 결혼여부, 직업이 유의한 변수로 나타났는데 미혼일수록, 직업이 전문직에 종사할수록 중저가 와인 구매횟수에 정(+)의 영향을 미치는 것으로 나타났다. 둘째, 사회심리적 특성으로는 태도와 지각된 행동통제감이 중저가 와인 구매횟수에 정(+)의 영향을 주는 것으로 나타났다. 셋째, 구매관련 특성인 사전지식, 그리고 즐거움 동기가 중저가 와인 구매횟수에 정(+)의 영향을 미치는 것으로 나타났다. 마지막으로 본 연구결과의 경영관리적 시사점과 연구의 한계점에 대해 토의하였다. 시사점 및 연구의 한계점: 이 연구결과는 와인 소비자의 의사결정, 구매행동에 대한 이론적, 실무적 이해를 도모했다는 것에 의의가 있다. 또한 이 연구는 기존 와인 소비자의 구매행동에 대한 선행연구에서 설명변수, 연구모델을 더욱 확장했다는 것에 학술적인 의의가 있다. 그러나 이 연구는 중저가 와인만을 대상으로 했으므로 연구결과의 일반화에 한계가 있을 수 있다. 향후 연구방향: 향후에는 더욱 다양한 가격대의 와인과 다양한 종류에 따른 소비자의 구매행동을 분석할 필요가 있을 것이다. 특히 와인 소비자를 세분화하여 소비자의 구매행동 특성을 분석할 필요가 있을 것이다. 공헌점: 이 연구는 중저가 와인 소비자의 구매행동 특성만을 분석하였으며, 분석모델로 구매횟수(6개월, 1년)에 따른 가산자료모형을 이용하였다. 이 연구는 특히 다양한 가격대의 와인 중에서도 가장 대중적인 중저가 와인 구매행동만을 분석하여 와인 소비자 구매행동에 대한 이해를 돕고자 하였다.

Purpose: The purpose of this study is to estimate the purchasing behavior of wine consumers using the count data model. Design/methodology/approach: To this end, the Truncated Poisson(TPOIS) Model and Truncated Negative Binomial(TNB) Model were used because the dependent variable of this study was zero-truncated count data . In order to understand the characteristics of wine purchasing behavior, the constructs of the Theory of Planned Behavior(TPB), prior knowledge and motivation were also included as independent variables in the model. Furthermore, the time periods of the dependent variable(frequency of purchasing wine) were specified with two categories(past 6 months and one year) because previous studies using the count data model only measured the dependent variable in a single period. Results/findings: The result of this study indicate that the TNB Model having the dependent variable with ‘the frequency of purchasing wine in the past six months’ was preferred over other models. The results also reveal that single, profession, attitude, perceived behavioral control, prior knowledge and motivation of pleasure had significantly positive effects on the purchasing behavior of wine. Research implications/limitations: The findings of this study reflect the increased awareness and interest in the decision-making of consumer behavior for wine in Korea. This study may be one of the forerunner studies that describe the decision-making of consumer behavior for wine by applying the count data model. Future work/research: Comparing each segment with different wine consumers by various market characteristics is likely to strengthen the current model(TPOIS & TNB). Originality/value: This research framework of the TPOIS & TNB models provides practical insights into the wine industry as well as theoretical implications for academics

This study is conducted to achieve the following three objectives. First, it is to determine the attributes for wine selection. Second, the wine demand is estimated for each market segment. Data are collected through the surveys of general public who have experiences in drinking wine. A convenience sampling method is used to select respondents in Seoul and the metropolitan area. A total of 500 questionnaires are distributed, 436 are used for final analysis. Analyses are performed on the data by using descriptive statistics, factor analysis, truncated Poisson model method. The findings of the study are as follows. First, the determining attributes for wine selection are extracted from the analysis. Second, Each group is named and the characteristics are described. The importance of each attribute is rated differently among the groups. Lastly, the factors affecting the wine demand are identified and the demand. The important factors identified, to name a few, are gender, income, education, and lifestyle for positive influence, and the negative influence factors are job and age. These findings may provide managerial implications for marketing plans for wine distributors.

본 연구에서는 ‘2018년도 부산지역 인력 및 훈련 수요조사’에서 부산지역 기업의 구인인원에 대한 결정요인을 분석하고자 한다. 일반적으로 구인인원은 0을 포함한 자연수를 가지는 계수 데이터이므로, 일반적인 선형회귀모형을 적용하기에는 정규성, 등분산성, 독립성 등에서 많은 문제점이 있다. 이런 문제를 극복하기 위해서 계수 데이터 모형인 포아송 회귀, 영과잉 포아송 회귀, 음이항 회귀 및 영과잉 음이항 회귀모형 등 4개의 모형을 적합하여 최적 모형을 선택한다. 분석의 주요결과를 요약하면 다음과 같다. 첫째, 모형 선택에 대한 가설검정 결과, 영과잉 음이항 회귀모형이 최적의 모형으로 선택되었다. 둘째, 분석결과 부산지역 기업들 가운데 비 IT 서비스업과 기타업종일수록, 기업규모가 클수록, 파견사업체나 용역(도급)업체일수록, 산업단지 이외 지역일수록, 구인인원이 유의하게 높게 나타났다. 셋째, 4차산업혁명 인지도 여부는 인지기업일수록 구인인원이 많으나 통계적으로 유의하지 않은 것으로 나타났다. 마지막으로, 이러한 분석결과는 부산지역 기업 가운데 제조업 및 산업단지 입주기업의 일자리창출에 대한 정책대안의 모색이 필요함을 시사한다.

In this paper, the determinants for volume of company recruitment in Busan area is analysed. Since volume of recruitment is count data which take natural number including zero values, linear regression model becomes inappropriate. Therefore, four models such as Poisson regression model, zero inflated Poisson regression model, negative binomial regression model and zero inflated negative binomial regression model are tried to fit count data. Major results are the followings. A zero inflated negative binomial regression model is selected to be the best model. Second, the volume of recruitment in the Busan area is significantly higher as the service industry, the larger the enterprise, the dispatching business, the service provider, the area outside the industrial complex. The recognition of the fourth industrial revolution was not statistically significant, although the number of recruits was higher in the cognitive firms. Finally, this analysis suggests that it is necessary to find policy alternatives for job creation in manufacturing and industrial complex enterprises in Busan.

The aim of this study was to examine the determinants of visit demand (VD) for Lake Osan so as to provide fundamental information for the improvement of waterfront area. Using 302 survey datum collected from local residents who had recently visited the site, the authors empirically tested the truncated negative binominal model using five factors derived from factor analysis, such as ‘Management operation (MO)', ‘Stream nature (SN)', ‘Recreational facility (RF)', ‘Convenience facility (CF)', and ‘Easy accessibility (EA)'. The main results are as follows. SN strongly had a positive impact on VD whereas EA demonstrated a negative effect, and furthermore it was revealed that the number of visits by married people was significantly more than unmarried ones. MO, RF and CF, meanwhile, were not significant on VD. In conclusion, we had a discussion about practical implications based on analysis results.

This study aims to develop the traffic accident models by circular intersection type using count data model. The number of accident, the number of fatal and injured persons(FSI), and EPDO are calculated from the traffic accident data of TAAS. The circular intersection accident models are developed through Poisson and negative binomial regression analysis. The main results of this study are as follows. First, the null hypotheses that there are differences in the number of traffic accidents, FSI and EPDO by type of circular intersections are rejected. Second, the scale of intersection(median, large), number of approach road, mean width and length of exit road, area of the circulating roadway and central island are selected as factors influencing the number of traffic accidents, FSI and EPDO in rotary. Third, the scale of intersection(median), guide signs(limited speed, direction, roundabout), number of approach road, entry angle, area of the intersection and central island are adopted as factors influencing the number of traffic accidents, FSI and EPDO in roundabout. Finally, transferring from rotary to roundabout could be expected to make the accident decrease.