Earticle

동일한 총 반응조 체적 및 운전조건에서 MLE와 Bardenpho 공법을 EQPS를 이용한 시뮬레이션을 통해 최종방류수 암 모니아성 질소 및 총 질소 농도를 파악하였다. 이차침전전지는 Gould Type I을 사용하는 것으로 간주해서 모든 시뮬레이 션에 있어서 생물반응조 MLSS 농도는 3,500 mg/L 근처를 가지도록 설정하였다. MLE공법 시뮬레이션에 있어서 무산소 조와 호기조 개수를 각각 1, 2 및 4개가 되도록 하였고, Bardenpho공법에서는 각 호기조 및 무산소 반응조 개수를 1과 2 개로 하였다. 시뮬레이션을 통해 반응조 개수가 증가할수록 질산화와 탈질 반응의 효율이 증가하는 것으로 나타났으며, 동일한 반응조 개수에서는 Bardenpho공법의 질산화 탈질효율이 MLE공법에 비해 우수한 것으로 나타났다.

Effluent nitrogens from both MLE and Bardenpho processes with same total reactor volume were characterizes by EQPS under same operation conditions. At simulation, MLSS concentrations at bioreactors were adjusted as approximately 3,500 mg/L which Gould Type I clarifier was assumed to be used for secondary clarifier. Both MLE and Bardenpho processes have equal total reactor volume. For MLE process simulation, the number of anoxic and oxic reactors are increased as 1, 2 and 4, respectively. For Bardenpho process, the number of each anoxic and oxic reactor is increased as 1 and 2, respectively. Through simulation, it is found that nitrification and denitrification efficiencies are increased as number of reactor is increased. Also, Bardenpho process shows higher nitrification and denitrification efficiencies than MLE process at the same number of reactors.

We propose a mean value function for software failures in NHPP software reliability model. And we deal with the maximum likelihood estimation and the least squares estimation in the proposed mean value function. The explicit mean value function solution for the proposed model is presented by MLE and LSE in two data sets. The values of SSE and MSE is presented in two data sets by MLE and LSE. We compare the predicted number of faults with the actual two data sets using the proposed mean value function.

정부는 2014년 우리나라의 일인당 GDP를 $28,160로 발표했다. 이는 처음으로 지난 10년만에 3만 달러를 육박하는 수치이다. 이는 경제선진국이란 관점에서 최근 사회적으로 양적 성장에 반하여 부의 양극화를 지적하는 목소리가 높고 불만의 소리가 큰 것은 우리사회의 부의 불평등구조를 대변하는 것이 아닌가 한다. 그래서 본 연구는 저 출산 고령화를 치닫고 있는 인구구조에서 여성의 경제활동은 부의 양극화 해소와 국민소득의 불평등도 해소에 어느 정도 기여하는지의 여부를 확인해보고자 한다. 이를 위해 경제학 일반의 소득 불균형 정도를 나타내는 지니계수를 종속변수로 하고 정부복지지출, 민간 복지지출, 여성의 경제 활동참가율, GDP, 질적 변수인 외환위기 전후 등 5개를 독립변수로 하여 다중회귀분석을 시도 한 결과 민간의 복지지출확대가 0.05% 유의 수준에서 통계적으로 유의한 한 음(-)의 관계를 보이나, 여성의 경제활동 참가는 0.05% 유의 수준에서 통계적으로 유의한 한 정(+)의 상관관계를 보이는 이례적인 결과가 나왔다. 즉, 여성경제활동참가율이 높을수록 지니계수도 높아지는 결과를 나타냈으나 이러한 결과는 여성이 경제활동에 참가함으로써 기회비용도 같이 증가하는 이유로 분석한다. 주목할 만한 것은 연구 분석 모형에서 이 변수의 계수 값이 다른 변수들에 비해 가장 큰 기울기를 나타냄을 보임으로써 여성의 경제활동 참가율의 중요성을 부각되는 의의가 있을 것으로 평가된다.

The government has announced a GDP per capita of the country in 2014 to $ 28,160. It is the first year that almost $30,000 was achieved for the last 10 years. Despite of the quantitative growth in recent years, according to bipoarization of wealth, social complaints against the economic inequality have been increased. Under the demographic structure of aging population and the low birth rates, it is examined the extent that economic activities of women can contribute to reducing income inequality and bipolarization of wealth. Welfare as the dependent variable and the Gini coefficient represents a disproportionate level of government in economics ordinary income for this spending, private social expenditure, women’s labor force participation rate, GDP, qualitative variables, such as before and after the financial crisis by five independent variables multiple regression statistically significant negative results conducted at the significance level of welfare spending in the private sector is 0.05% analysis negative(-) but the relationship between the labor force participation of women was statistically significant correlation between the positive(+) at the signifinance level of 0.05% results came out showing an unusual relationship, In other words, the higher the female labor force participation rate but actually showed a higher Gini coefficient also results. The results will be analyzed women as a reason to increase the opportunity cost by participating in economic activities. It is estimated that there are notable significance that emphasize the importance of labor force participation rate of women by showing the slope represents the largest study analyzing the model compared to other variables in the coefficients of the variables.

본 논문에서는 근거리 무선 통신 기술인 Zigbee를 활용하고, RSSI 핸드오버 기법을 통하여 보다 정확하고 효율적인 실내 위치 인식 시스템을 제안한다. RSSI(Received Signal Strength Indication) 값을 이용한 실내 위치 인식방식에는 여러 가지가 있는데 가장 대표적인 방법으로는 삼각측량법을 이용하는 경우이다. 하지만 단순히 삼각측량법으로 측정된 RSSI 값은 여러 가지 환경적 요인으로 인해 불규칙하게 되는데, 본 논문에서는 보다 정확한 실내위치 인식을 위하여 RSSI 값의 채널 간 이동, 즉 핸드오버 (Handover)를 이용하여 각 채널간의 간섭을 줄여 안정적인 RSSI 값을 측정하는 방식을 제안한다. 이때 핸드오버는 고정 노드가 아닌 이동노드의 이동식 셀을 사용한 핸드오버를 사용하여 위치 인식에 필요한 세 개의 고정노드만을 선택하는 방법을 사용하였다. MLE(Maximum Likelihood Estimation) 방법을 통해 이동 노드와 고정노드의 거리를 측정하는 동시에 삼각 측량법을 이용하여 이동노드의 좌표를 구하여 보다 정확한 실내 위치를 인식할 수 있는 시스템을 제안하고 구현하여 검증하였다.

In this paper, an accurate and efficient indoor location-aware system is proposed through RSSI(Received Signal Strength Indication) handover technique and using the Zigbee, short-range wireless communication protocol. Among the solutions of several indoor location-aware systems with the use of RSSI value, the using of triangulation measuring method is the most popular. However, due to various environmental factors, the measured RSSI values with the triangulation measuring method are unstable. In this paper, for accurate indoor location-aware system, we propose more stable RSSI value measurement method using the RSSI value transfer between channels, that is, the handover approach. The handover of the mobile node instead of the fixed nodes by using a removable cell handover with the need for location-aware how to select only three fixed nodes were used. Through the finding of the mobile node's coordinates using triangulation measuring method in addition to the measuring the distance between the fixed node and the mobile node using the MLE (Maximum Likelihood Estimation) technique, more accurate indoor location-aware system is proposed, implemented and proved.

최근 스마트교육을 활성화를 위한 스마트러닝 (Smart Learning: s-learning)이 확산됨에 따라 다양한 교육용 콘텐츠가 개발되고 있다. 본 연구의 목적은 스마트 애플리케이션의 기능을 분석하여 살펴보고, ADHD (Attention Deficit Hyperactivity Disorder) 아동의 특성을 고려하여 중재학습경험이론의 이론을 기초로 한 스마트 교육용 시스템을 설계 및 제작하는 것이다. 본 시스템은 ADHD 아동의 자아효능감 향상을 위해 MLE (Mediated Learning Experience: MLE)기반의 시스템을 설계하였고 ADHD 아동에게 적용한 결과는 다음과 같다. 첫째, ADHD 아동의 학습 흥미도를 높여 주의집중력 향상 및 학습 능력의 증진이 가능하다. 둘째, 성공적인 학습 경험을 토대로 자신감과 자긍심을 높여 자아효능감을 증진시킨다. 셋째, 메타인지의 발달에 따른 인지구조의 변화로 학업 성취도가 향상된다.

In this paper, a smart education system is developed and implemented for ADHD students. Usually, ADHD students lack of self efficacy. Self efficacy is very important factor for improving social adaptability and learning effect of ADHD students. In the proposed system, MLE concept is adapted. MLE concept is used to improve self efficacy of ADHD students. The purpose of the proposed system is to help ADHD students have high study capability. The proposed system is applied an ADHD student. The following results are obtained. First, the system can improve study interests. In turn, the system is helpful to improve concentration and learning effect. Second, based on successful study experience, self efficacy is improved. Third, study achievement is improved by changing cognitive structure that is due to development of meta-cognition.

Background/Objectives: This study was designed to compare the Varimax analysis of Principal Component Analysis(PCA) with the Exploratory Factor Analysis(EFA) method using Maximum Likelihood Estimation(MLE) and oblique rotation(Direct Oblimin) for the multicultural experience of university students. multicultural experience variables of college students were analyzed as two factors through Maximum Likelihood Estimation(MLE). Methods/Statistical analysis Q1, Q2, Q3 are multicultural direct experiences, and Q4, Q5, Q6, Q7 are multicultural indirect experiences. Findings: As a result of this study, it is suggested to use Maximum Likelihood Estimation(MLE) and oblique rotation(Direct Oblimin). As a prerequisite for applying the Maximum Likelihood Estimation(MLE), it is thought that Improvements/Applications it is necessary to increase the accuracy of the response rate by using the face-toface interview method when conducting a survey.

Based on the CDM model, this paper empirically examines the links of R&D, innovation, and labor productivity using Korean Community Innovation Survey (KCIS) for manufacturing firms over the period 2013 to 2015. Estimating the model, I attempt to alleviate two major econometric problems such as endogeneity for the variables and sample selection biases caused by distribution problems of R&D spending and technological innovations. To do this, I control the former bias by employing sequential instrumental variable estimation procedures, and the latter by using both the Generalized Tobit model and the bivariate probit estimation technique. Main empirical findings in the paper can be summarized as follows. First of all, this paper found that the separate estimation of the structural CDM model should lead to biased results due to ignoring sample selection and endogeneity problems. This evidence confirms validity of using botht the bivariate probit model and the type II Tobit model. Second, the results show that R&D expenditure should be positively associated with enhancement of product innovation, rather than process innovation. This implies that government R&D support can have a stronger effect on encouraging product innovation through an increase in R&D investment in SMEs. Third, the innovation protection methods, both legal and strategic, are found to have a positive effect on R&D investment but no effect on R&D intensity. In the innovation output models, on the other hand, only strategic protection method can have a positive effect on both product and process innovations. Fourth, the firm size proxied by the number of workers is found to be negatively related to the R&D investment. As the size of firms grows, the probability of R&D investment is higher, but per worker R&D investment is smaller. This evidence supports Scherer’s hypothesis that innovation activities are more active of small-medium firms in the competitive markets, rather than in the monopoly market suggested by Schumpeter. Fifth, the results show that information flow from suppliers and customers plays an important role as source of knowledge for invention and innovation. In the knowledge production function or the innovation output model, the flow has a positive effect on both the product and process innovation. However, information flow from competitors is found to affect process innovation, but it is not useful to develop new products. Next, the estimated coefficients for government financial support are found to be statistically significant in both the R&D investment decision and the R&D intensity models. This implies that the government financial support can have crowding-in effects on R&D investment, rather than crowding-out effects, and thus it could lead to an enhancement of R&D investment in SMEs. Finally, the results show that the estimated coefficients of the predicted values for product innovation are statistically significant in the labor productivity model, implying that product innovation should have a positive effect on labor productivity and thus eventually affect economic growth positively. This results also confirms existence of endogeneity problem with product innovation, not with process innovation.

Noise power spectral density (PSD) estimation is a crucial part of speech enhancement system due to its contributory effect on the quality of the noise reduced speech. A novel estimation method for color noise PSD on the basis of an assumption of generalized Gamma distribution and maximum a posteriori (MAP) criterion is proposed. In the experiment, generalized Gamma PDF which is a natural extension of the Gaussian modeling of a non-white components distribution is found best fitting in four types of color noises compared with Laplace, Rayleigh distributions. After that MAP noise estimators based on the reported generalized Gamma PDF models are competed with Minimum Statistics (MS), minimum mean square error (MMSE) based PSD estimation and Maximum Likelihood estimation (MLE) noise tracking methods in evaluations. The performance of the proposed noise estimations are good as demonstrated by log error, segmental SNR and PESQ measures when they are integrated with the speech enhancement technique.

본 연구의 목적은 G-DINA(generalized deterministic input, noisy "and" gate)모형의 통계적 속성을 더 잘 이해하기위해서, 몇몇 요인들이 인지요소숙달분류 정확성에 어떠한 영향을 미치는가에 대해 알아보는데 있다. 인지요소숙달 분류의 정확성에 영향을 미치는 요인으로 인지요소의 수, 표본크기, 검사 길이를 달리 하였을 때, 인지요소숙달의 분류 정확성의 변화에 대해 제시했다. 이를 위해 본 연구에서는 모의실험을 통하여 다양한 조건의 데이터를 생성하고, 각 응시생에게 할당된 참인지요소와 ML방법으로 추정된 인지요소간의 분류 정확 비율과 카파계수를 통하여, 인지요소숙달 분류 정확성을 갖는 검사조건에 대하여 밝히고자 한다. 연구결과, G-DINA 모형에서 인지요소의 수가 5~7개 일 때, 표본크기는500명 이상, 검사 길이는 30문항 이상이 될 때, 인지요소숙달의 분류 정확성이 높은 것으로 나타났다. 본 연구결과는 G-DINA 모형을 사용해서 자료를 분석 할 때, 인지요소숙달의 분류 정확성을 갖기 위해서 연구자가 고려해야할 검사조건들에 대한 가이드라인으로 활용될 수 있다.

The purpose of this study was to better understand the statistical properties of the G-DINA(gemeralized deterministic input, noisy "and" gate) model, the impact of several fators on the classification accuracy was investigated. Factors affecting the classification accuracy is the number of attribute, sample size, test length. results for classification accuracy are reported using Cohen's Kappa and percent of exact agreements. For this study is to generate data of various conditions through simulation. For each condition was to examine the classification consistencies between true examinee skill patterns and estimated skill patterns (MLE classified examinees). The results showed that the classification accuracy of the G-DINA held for sample sizes(500 examinees), test lengths(30 items). when analyzed the data using the G-DINA, researchers will be able to be utilized as a guideline for classification accuracy.

The goal of automatic speech recognition (ASR) system is to accurately and efficiently convert a speech signal into a text message independent of the device, speaker or the environment. In general the speech signal is captured and pre-processed at front-end for feature extraction and evaluated at back-end using the Gaussian mixture hidden Markov model. In this statistical approach since the evaluation of Gaussian likelihoods dominate the total computational load, the appropriate selection of Gaussian mixtures is very important depending upon the amount of training data. As the small databases are available to train the Indian languages ASR system, the higher range of Gaussian mixtures (i.e. 64 and above), normally used for European languages, cannot be applied for them. This paper reviews the statistical framework and presents an iterative procedure to select an optimum number of Gaussian mixtures that exhibits maximum accuracy in the context of Hindi speech recognition system.

본 연구에서는 생활폐기물 매립지에서 발생되는 침출수에 함유된 고농도 $NH_4{^+}$-N 및 T-N의 적정처리를 위하여 pilot 규모의 MLE 공정을 이용하여 생물학적 질산화/탈질을 실시한 결과 $NH_4{^+}$-N온 99%이상, T-N은 88%정도의 처리효율을 얻을 수 있었으며, BOD/$NH_4{^+}$-N비를 약 3.5로 유지할 경우 T-N온 $0.09kgN{\cdot}m^{-3}{\cdot}day^{-l}$정도 제거되는 것으로 나타났다. 또한, 질산화/탈질과정에서 $NH_4{^+}$-N 산화량에 따른 알카리도 소비량은 평균 $3.4{\sim}3.5kgAlk{\cdot}kgNH_4{^+}-N^{-1}$정도로 나타났으며, 외부탄소원으로 투입된 메탄올에 대한 탈질균의 적용 기간은 약 20일 정도가 필요한 것으로 나타났다. 또한, 질산화조에 유동상 메디아를 투입한 결과 투입하지 않았을 때에 비하여 $SV_{30}$가 2배 정도 양호하게 나타났다.

This study was carried out to investigate the treatability of landfill leachate having high concentration of ammonium nitrogen with/without the circulation of media in pilot-scale($48m^3basis$) process. Total nitrogen removal efficiency was relatively increased in the media added process (influent ; $1.230{\sim}2,000mg{\cdot}l^{-1}$, effluent ; $120{\sim}250mg{\cdot}l^{-1}$) compared with the control process. The difference of nitrogen removal efficiency between these processes may be due to that stable growth of nitrifiers attached to the media could be achieved 99.3% of ammonium-nitrogen removal efficiency(without ; 98.2%) and 88.5% of total nitrogen removal efficiency(without ; 85.8%) were shown in media added process, respectively. Also, optimum BOD/ $NH_4{^+}$-N ratio was relatively decreased in the media process compared with the control process. Sludge settleability, on the other hand. was shown better in media added process than in control process. This outstanding sludge settleability in the media added process indicates the compatibility of media(zeolite) to the microorganism and the possibility of using media of biofilm process.

In this study, green house gas emission from a wastewater treatment plant (WWTP) was estimated as a part of a project to build a national greenhouse gas inventory. Gas from the WWTP under study, which was employing an MLE process was collected with a dynamic flux chamber, and methane(CH4) and nitrou oxide(N2O) in the gas were analyzed with a NDIR. The emission rates of CH4 and N2O from a primary sedimentation tank were 9,876.87 μg/min and 2,352 μg/min, respectively. Emission factors of CH4 and N2O of the plant were 0.1089 kgCH4/kgBOD5 and 0.0414 kgN2O/kgTN, respectively.

본 연구에서는 DINA 모형 하에서 피험자 능력 모수를 추정할 수 있는 방법인 MLE, MAP, EAP 각각의 분류 정확도 및 상호 간의 합치도를 모의실험을 통하여 조사하였다. 선행연구와의 차별성을 확보하기 위하여 문항의 질, 검사 길이, 능력 정도, 인지요소 간 상관 정도, 그리고 검사 성격 등의 다양한 모의실험 요인을 고려하였다. 연구 결과, 개별 인지요소를 추정할 때의 정확성은 어느 방법을 사용하여도 모든 조건에서 약 75% 이상인 것으로 나타났다. 하지만 전체 인지요소를 모두 함께 정확하게 추정하는 것은 상대적으로 어려운 것으로 나타났으며 특히 문항의 질이 좋지 않고 문항 수가 적을 때에는 50% 초반까지 분류 정확도가 하락하였다. 능력추정 방법 간 결과를 비교해 보면, 개별 인지요소를 추정할 때에는 EAP를 사용하는 것이 바람직하지만 전체 인지요소 모두를 정확하게 추정하는 것이 더 중요하다면 MLE/MAP를 선택하는 것이 나았다. 다음, 검사별 Q-matrix 구성에 따라서 능력 추정의 정확성에 영향을 미치는 것을 알 수 있었는데 검사 문항 수에 비하여 상대적으로 많은 인지요소를 고려할 경우 분류 정확도가 낮아짐을 확인하였다. 문항 수가 많을수록 그리고 문항의 질이 좋을수록 분류 정확성이 높아짐을 알 수 있었다. 흥미롭게도 특정 인지요소만을 측정하는 문항이 5문항 이상인 경우 문항의 질이 다소 하락하여도 높은 분류 정확도를 구할 수 있었다. 또한 인지요소 간의 상관 정도는 능력 추정의 정확도와 무관한 것으로 나타났다.

In this simulation study, the classification and agreement of MLE, MAP, and EAP which are methods to estimate examinees’ ability parameters under the DINA model. For the purpose of ensuring differentiation from the previous studies, various simulation factors were considered such as quality of test items, test lengths, ability distributions, correlations among cognitive attributes, and natures of test. The results showed that the accuracy of estimating each individual cognitive attribute was higher than about 75% in all conditions using any of three methods. It was relatively difficult, however, to accurately estimate all the cognitive attributes together. Especially, when the number of test items is small, the classification accuracy has decreased to the early 50%. It was better to use EAP when estimating individual cognitive attribute. If it was more important to accurately estimate mastery pattern, however, it is better to choose MLE/MAP. Next, the Q-matrix configuration appeared to affect the accuracy of ability estimation. As the number of cognitive attributes came to be relatively larger compared to the number of related test items, the classification accuracy tended to be lower. The higher the number of items and the better the quality of items, the higher the classification accuracy. Interestingly, if the number of items that measure a specific cognitive attribute were more than 5, the ability estimation accuracy could be kept relatively high even though the quality of items dropped somewhat. Also, the amounts of correlations among cognitive attributes appeared not to be related to the accuracy of ability estimation.

컴퓨터시스템은 여러 가지 복잡하고 민감한 시스템을 제어하는데 광범위하게 쓰이고 있다. 최근에 와서는 운영시스템, 제어프로그램, 적용프로그램과 같은 여러 가지 소프트웨어 시스템이 더욱더 복잡화 및 대형화되고 있기 때문에 신뢰성이 높은 소프트웨어 시스템을 개발하는 일이 매우 중요하며, 소프트웨어 제품개발에 있어서 소프트웨어의 신뢰도가 핵심사항이다. 1970년대 이후 소프트웨어의 신뢰성을 향상시키기 위한 여러 가지 소프트웨어의 신뢰도 모델이 제시되고 검토되었으며, 특히, 소프트웨어 개발 후 테스트단계에 적용하는 신뢰도를 추정하고 예측하는 모델이 많이 개발되었다. 소프트웨어가 주어진 시간간격동안 고장이 발생하지 않을 확률 즉, 신뢰도는 소프트웨어의 테스트과정을 계속해서 반복 및 수정하면 더욱 더 증가된다. 그러한 결함검출현상을 설명해주는 소프트웨어 신뢰도 모델을 소프트웨어 신뢰도 성장 모델(SRGM)이라 한다.

Two estimators for the scale (${\delta}$) and shape (${\beta}$) parameters and percentiles of the Weibull distribution were compared. These estimators are maximum likelihood estimator (MLE) and the best linear unbiased estimator (BLUE). The performance of these estimators are compared by mean square error and studied in complete and type II censored samples of size 10 and 25. The overall performance of the MLE was similar to that of the BLUE.

Maximum likelihood estimate(MLE) is obtained from the partial log-likelihood function for the cell probabilities of two way incomplete contingency tables proposed by Chen and Fienberg(1974). The partial log-likelihood function is modified by adding lagrangian multiplier that constraints can be incorporated with. Variances of MLE estimators of population proportions are derived from the matrix of second derivatives of the loglikelihood with respect to cell probabilities. Simulation results, when data are missing at random, reveal that Complete-case(CC) analysis produces biased estimates of joint probabilities under MAR and less efficient than either MLE or MI. MLE and MI provides consistent results under either the MAR situation. MLE provides more efficient estimates of population proportions than either multiple imputation(MI) based on data augmentation or complete case analysis. The standard errors of MLE from the proposed method using lagrangian multiplier are valid and have less variation than the standard errors from MI and CC.

최근 음성 인식 시스템의 성능 향상을 위해 화자 적응 (speaker adaptation)에 대한 연구가 활발히 진행되고 있다. HMM 기반 인식 시스템의 모델 파라미터를 수정하는 화자 적응의 경우, MAP방법과 MLLR 방법에 대한 연구가 주류를 이루고 있다. 두 방법은 adaptation data의 양에 따라서 서로 다른 성능을 보인다. 본 논문에서는 기존 두 방법을 Maximum-likelihood Estimation(MLE)를 이용하여 화자 적응을 수행하는 방법을 제안한다. 제안한 방법을 KAIST 통신연구실에서 구축한 한국어 도시이름 500단어 인식 시스템에 적용하여 adaptation data의 양에 상관없이 항상 높은 성능을 나타냈으며, 기존의 방법에 대해서 최고 4.37%의 인식률 향상을 보였다.

MLE and UMVUE of the right-tail probability in a Levy distribution will be considered, and hence the UMVUE is more efficient in a sense of simulated MSE than the MLE of the right-tail probability.

It is known that the maximum likelihood method does not provide explicit estimator for the scale parameter of the Weibull distribution based on Type-II censored samples. In this paper we provide an approximate maximum likelihood estimator (AMLE) of the scale parameter of the Weibull distribution with Type-II censoring. We obtain the asymptotic variance and simulate the values of the bias and the variance of this estimator based on 3000 Monte Carlo runs for n = 10(10)30 and r,s = 0(1)4. We also simulate the absolute biases of the MLE and the proposed AMLE for complete samples. It is found that the absolute bias of the AMLE is smaller than the absolute bias of the MLE.