• 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.

• The KIPS Transactions:PartC
• /
• v.19C no.4
• /
• pp.239-246
• /
• 2012

The function of location estimation in WPAN has been studied and specified on the ultra wide band optionally. But the devices based on CSS(Chirp Spread Spectrum) specification has been used widely in the market because of its functionality, cheapness and support of development. As the CSS device uses 2.4GHz for a carrier frequency and the sampling frequency is lower than that of the UWB, the resolution of a timestamp is very coarse. Then actually the error of a measured distance is very large about 30cm~1m at 10 m depart. And the location error in ($10m{\times}10m$) environment is known as about 1m~2m. So for some applications which require more accurate location information, it is very natural and important to develop a sophisticated post processing algorithm after distance measurements. In this paper, we have studied extended Kalman filter with the frequency offsets of anchor nodes, and proposed a novel algorithm frequency offset compensated extended Kalman filter. The frequency offsets are composed with a variable as a common frequency offset and constants as individual frequency offsets. The proposed algorithm shows that the accurate location estimation, less than 10cm distance error, with CSS WPAN nodes is possible practically.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.15 no.6
• /
• pp.655-660
• /
• 2005

This paper studied the feature parameters less affected by the emotional variation for the development of the robust speech recognition technologies. For this purpose, the effect of emotional variation on the speech recognition system and robust feature parameters of speech recognition system were studied using speech database containing various emotions. In this study, LPC cepstral coefficient, met-cepstral coefficient, root-cepstral coefficient, PLP coefficient, RASTA met-cepstral coefficient were used as a feature parameters. And CMS and SBR method were used as a signal bias removal techniques. Experimental results showed that the HMM based speaker independent word recognizer using RASTA met-cepstral coefficient :md its derivatives and CMS as a signal bias removal showed the best performance of $7.05\%$ word error rate. This corresponds to about a $52\%$ word error reduction as compare to the performance of baseline system using met - cepstral coefficient.

• Journal of Korea Water Resources Association
• /
• v.36 no.2
• /
• pp.315-324
• /
• 2003

Stream flow data was analyzed for determining the lowflow which is the standard for river maintenance flow. Lowflow quantiles were estimated based on the parametric and nonparametric methods and two methods were compared by Monte Carlo simulation study. As the results of the parametric method, three probability distributions such as gamma-2, lognormal-2 and Weibull-2, are selected as appropriate models for stream flow data of 13 stations in Han River Basins. According to simulation results, relative bias (RBIAS) and relative root mean square error (RRMSE) of the lowflow quantiles are the smallest when the applied and population models are the same. The fame statistical properties from the nonparametric models are good within the interpolation range. Among 7 bandwidth selectors used in this study, the RRMSEs of the Park and Marron method (PM) are the smallest while those of the Shoaler and Jones method (SJ) are the largest.

• Korea journal of population studies
• /
• v.11 no.1
• /
• pp.1-13
• /
• 1988

This study examines the sampling bias that may have resulted from the large number of missing observations. Despite well-designed and reliable sampling procedures, the observed sample values in DSFH(Demographic Survey on Changes in Family and Household Structure, Japan) included many missing observations. The head administerd survey method of DSFH resulted in a large number of missing observations regarding characteristics of elderly non-head parents and their children. In addition, the response probability of a particular item in DSFH significantly differs by characteristics of elderly parents and their children. Furthermore, missing observations of many items occurred simultaneously. This complex pattern of missing observations critically limits the ability to produce an unbiased analysis. First, the large number of missing observations is likely to cause a misleading estimate of the standard error. Even worse, the possible dependency of missing observations on their latent values is likely to produce biased estimates of covariates. Two models are employed to solve the possible inference biases. First, EM algorithm is used to infer the missing values based on the knowledge of the association between the observed values and other covariates. Second, a selection model was employed given the suspicion that the probability of missing observations of proximity depends on its unobserved outcome.

• The Korean Journal of Applied Statistics
• /
• v.30 no.3
• /
• pp.453-462
• /
• 2017

Double propensity score adjustment is an analytic solution to address bias due to incomplete matching. However, it is difficult to estimate the standard error of the estimated treatment effect when using double propensity score adjustment. In this study, we propose two bootstrap methods to estimate the standard error. The first is a simple bootstrap method that involves drawing bootstrap samples from the matched sample using the propensity score as well as estimating the standard error from the bootstrapped samples. The second is a complex bootstrap method that draws bootstrap samples first from the original sample and then applies the propensity score matching to each bootstrapped sample. We examined the performances of the two methods using simulations under various scenarios. The estimates of standard error using the complex bootstrap were closer to the empirical standard error than those using the simple bootstrap. The simple bootstrap methods tended to underestimate. In addition, the coverage rates of a 95% confidence interval using the complex bootstrap were closer to the advertised rate of 0.95. We applied the two methods to a real data example and found also that the estimate of the standard error using the simple bootstrap was smaller than that using the complex bootstrap.

• Journal of the Korean earth science society
• /
• v.34 no.7
• /
• pp.626-632
• /
• 2013

In this study the GPS Precipitable Water Vapor (PWV) was derived and evaluated by a radiosode measure during the winter intensive observation in Gangneung site from January 5 till February 29 in 2012. Bernise 5.0 software was used to derive the GPS data. GPS-derived PWV from Zero difference (GANG) and Single difference (GANG and DAEJ) was high variance in time and about 5 times the PWV of radiosonde. GPS post-processing has been performed from two additional IGS site (Xian Dao, Ibaraki-ken) in order to correct the absolute troposphere errors. As a result, the mean bias error (MBE) and root mean square error (RMSE) and correlation compared with radiosonde measure were 0.67 mm, 6.40 mm, and 0.93, respectively. In order to correct the relative troposphere errors from the altitudinal difference between the two GPS receivers, we calculated the GPS-derived PWV by adding the data of GPS that was installed in Gangneung-Wonju University near the Gangwon Regional Meteorological Administration. In the end, the improved result showed that MBE, RMSE and correlation in comparison with radiosonde measures were 0.61 mm, 5.79 mm, and 0.93, respectively.

• Proceedings of the Korean Statistical Society Conference
• /
• 2003.10a
• /
• pp.53-58
• /
• 2003

최근 경제 사회의 급속한 발전에 따라 개인의 활동분야가 다양해지고 개인비밀보호 인식이 커지면서 응답자들이 통계조사에 잘 협조하지 않는 경우가 늘어나고 있다. 따라서 대부분의 통계를 현장조사에 의존하여 생산하고 있는 통계청에서는 현장조사 결과물의 품질상태에 좀 더 관심을 가져야만 하게 되었다. 이러한 현장조사에 대한 풀질관리의 일환으로 현재 통계청에서는 통계별 조사대상으로부터 품질관리표본(Quality Control Sample)을 추출하고 이를 대상으로 리인터뷰를 실시함으로써 응답오차측정, 조사현장의 실터파악, 응답자 의견수렴 등에 활용하고 있다. 리인터뷰는 조사직원의 고의적인 자료조작 또는 보충교육 필요성 등 현장조사업무를 평가하거나 응답분산(simple response variation), 응답편의(response bias) 등을 산출하고, 이를 분석하는 모델을 이용하여 응답결과의 신뢰도를 분석하는데 목적을 두고 있다. 본 연구에서는 품질관리표본(QC Sample) 설계 및 추출, 리인터뷰 시나리오개발, CATI(Computer Assisted Telephone Interviewing)를 이용한 리인터뷰 기법 등을 통계청 사업체조사 모니터링 사례를 중심으로 설명하고 조사직원 특성별 응답오차 측정 및 비교, 정확성 항목에 대한 차이분석 등 격과에 대하여 논하고자 한다.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2017.05a
• /
• pp.2-2
• /
• 2017

수문분석이 있어 정확한 강우량 추정 및 강우 자료의 품질은 매우 중요한 요소이다. 유출분석의 기본 입력 자료인 만큼 홍수유출 결과에도 큰 영향을 미치게 되는데, 현재 하나의 확정적인 값으로 제공되는 레이더 강우 자료는 추정과정에서 많은 오차 및 불확실성을 포함하고 있다. 강우 자료의 불확실성은 기상현상의 예측능력 한계로 인한 것으로 관측지점에서의 발생 가능한 다양한 강우시나리오의 범위를 나타낸다. 본 연구에서는 임의의 값을 추정하는데 있어 하나의 값이 아닌 가능한 값들의 범위를 정의하거나 확률분포를 표현할 수 있는 확률론적인 방법을 이용하여 레이더 강우 앙상블을 생성하고자 하였다. 2012년 남강댐 유역에 발생한 태풍 '산바', '볼라벤'을 대상으로 자료간 오차 공분산을 고려하 2012년 남강댐 유역에 발생한 태풍 '산바', '볼라벤'을 대상으로 자료간 오차 공분산을 고려하여 강우 앙상블을 생성하였으며, 레이더 강우에 내포된 불확실성 정도를 정량적으로 제시하였다. 생성된 강우 앙상블은 레이더 강우의 전체적인 편의보정뿐만 아니라 지상강우의 패턴을 잘 모의하고 있는 것으로 나타났으며, 레이더에 의해 추정한 강우의 불확실성을 잘 표현하고 있는 것으로 확인되었다. 강우 앙상블 생성 방법은 발생 가능한 다양한 강우 시나리오를 제공할 수 있으며 홍수예경보와 같은 의사 결정에 유용한 정보를 제공할 수 있을 것으로 판단된다.

• Journal of Korean Society of Forest Science
• /
• v.100 no.1
• /
• pp.62-69
• /
• 2011

Six commonly used non-linear growth functions were fitted to individual tree height-dbh data of eight major tree species measured by the $5^{th}$ National Forest Inventory in Chungcheong province. A total of 2,681 trees were collected from permanent sample plots across Chungcheong province. The available data for each species were randomly splitted into two sets: the majority (90%) was used to estimate model parameters and the remaining data (10%) were reserved to validate the models. The performance of the models was compared and evaluated by $R^2$, RMSE, mean difference (MD), absolute mean difference (AMD) and mean difference(MD) for diameter classes. The combined data (100%) were used for final model fitting. The results showed that these six sigmoidal models were able to capture the height-diameter relationships and fit the data equally well, but produced different asymptote estimates. Sigmoidal growth models such as Chapman-Richards, Weibull functions provided the most satisfactory height predictions. The effect of model performance on stem volume estimation was also investigated. Tree volumes of different species were computed by the Forest Resources Evaluation and Prediction Program using observed range of diameter and the predicted tree total height from the six models. For trees with diameter less than 30 cm, the six height-dbh models produced very similar results for all species, while more differentiation among the models was observed for large-sized trees.