반복측정 ANOVA 모형의 이해와 적용

엄한주 성균관대학교 2001 스포츠과학논집 Vol.6 No.-

기록과 측정의 과학-스포츠 현장에서 컴퓨터로 실시간 기록.분석한다

엄한주,Eom, Han-Ju 한국과학기술단체총연합회 2007 과학과 기술 Vol.40 No.9

체육측정평가 : 구형성 가정 위반이 신뢰계수 (α) 추정에 미치는 영향

엄한주(EomHan-Joo) 한국체육학회 2000 한국체육학회지 Vol.39 No.4

신뢰계수(α)의 추정이 반복측정 ANOVA 모형을 통해서 이루어지며, ANOVA에서 제공되는 제곱평균은 구형성 가정 위반에 영향을 받는다는 사실에 근거하여 본 연구에서는 구형성 가정 위반이 추정된 신뢰계수의 표본분포에 미치는 영향을 연구하기 위해 다양한 실험조건을 선택하여 각각의 실험조건에서 모의실험(simulation study)을 통하여 추정된 신뢰계수를 비교하였다. 자료의 특성이 범주형 자료인 경우에 추정된 α 계수들은 연속변인에서 추정된 α 계수에 비해 현저하게 낮게 추정되었다. 특히, 범주형 자료의 범위가 2-3점 척도일 경우에는 약 20% 정도의 감소가 있었다. 그리고 ANOVA에 의해 추정되는 α 계수는 일관성은 있으나 전반적으로 하향추정되는 특성이 있으므로 사례수가 적은 경위(n<30)나 또는 신뢰계수가 .75미만인 경우, 특히 범주형 자료일 경우에는 불편중 추정 공식을 사용할 것은 제안하였다. 그리고 연속변인인 경우에는 구형성 가정이 심각하지 위반되지 않는 한( ε>.70), α 계수의 표본이론과 통계적 검증 방법은 항내성이 있으나 척도의 범위가 3개 미만인 범주형 자료인 경우 에는 α 계수의 표본이론이 부적합하다는 것을 제시하였다. The present study emplyed Monte Carlo procedure to investigate the effects of noncircularity and data categorization on reliability coefficient estimates. Computer program, FORTRAN, was developed to conduct a series of simulations under various sampling conditions. Five independent parameters were initially considered in the simulations: three level of population alpha, three level of epsilon values, three level of repeated measures, three sample sizes, and four types of data. The results obtained from each condition were based on 2,000 - 6,000 repeated analyses. Due to a large amount of data the results from various conditions were integrated and accordingly presented. It was shown that the categorization of continuous data resulted in a considerably smaller a than for the parent continuous data, especially for a 3-point or less scale. Furthermore, the estimation procedure based on ANOVA produced a consistent, but negatively biased estimate. This bias was more noticeable for a small sample size, coupled with a small population α value. Thus, the use of an unbiased estimate was recommended for this type of design. Noncircularity conditions did not introduce any bias to the mean of the estimates, but yielded more variable estimates, thus increasing the variability of sampling distribution of a. The sampling theory of α with continuous data was fairly robust to a moderate departure from circularity, but somewhat sensitive to a severe nancircularity. However, the sampling theory of α was not adequate for categorical data, especially for a 3-point or less scale as it failed to produce appropriate proportions beyond the theoretical tolerance intervals.

배구경기 분석프로그램의 입력정보 설정 및 기술수행의 효율성 분석: 프로배구 2018-2019 정규시즌 여자경기 중심으로

엄한주(Han Joo EOM),김정아(Jung A KIM),최형준(Hyongjun CHOI) 한국체육측정평가학회 2020 한국체육측정평가학회지 Vol.22 No.4

The purpose of this study was two-folds a) to define and clarify the input variables being programmed in Volleyball Information System and b) to analyze and compare the skill efficiency of transition play in KOVO women’s pro volleyball. In recent years, computer applications become popular in competitive sport fields and have been widely applied to various sports for official match information as well as for performance analysis. Volleyball was one of the frontiers that introduced Computer Information System in international official games. Several programs are currently available and being used in the competitive field. However, the definition and clarification of the input variables are not consistent from one to another. This is especially so when the variables generated from a system were used and analysed for the academic purposes. As such, the meaning and interpretation of the results were unclear and incomparable among similar researches. This study proposed a specific notational criteria of input variables and refined the meaning of outcome variables being used for performance analysis. In addition, the variables generated from a system in KOVO were analysed by hierarchical log-linear method as an demonstrative example in order to investigate the effects of teams and win/lose on the transition actions of ‘Receive-Attack’ process. The analytic procedure and interpretation of results were provided in details with some suggestions for those who might use multiway frequency data produced in various sport events in years to come.

요인분석 모형의 이해와 적용 : 주성분모형과 공통요인모형의 방법론적 비교

엄한주(Han Joo Eom) 한국체육측정평가학회 2001 한국체육측정평가학회지 Vol.3 No.1

축구기술에 대한 객관적 측정자료와 주관적 평가자료의 비교

엄한주 ( Han Joo Eom ),김목수 ( Mok Soo Kim ) 한국스포츠정책과학원(구 한국스포츠개발원) 2002 체육과학연구 Vol.13 No.2

본 연구는 축구 기초기술에 대한 객관적 측정자료와 지도자의 주관적 평가를 비교하기 위하여 축구경기의 주요 기술인 헤딩, 지그재그 드리블, 롱 킥, 드리블 슛, 슛, 센터링, 30m 드리블, 10m 왕복 드리블 등 8개의 측정항목을 선정하여 중(n=112), 고(n=133), 대학교 선수(n=114)들을 대상으로 축구기술을 측정하였다. 각 측정항목에 대한 지도자의 주관적 평가점수와 측정자료의 관련성을 상관 및 교차분석을 통하여 비교하였으며, 선수들의 종합경기력 평가점수와 가장 관련이 깊은 기술요인을 선별하기 위하여 객관적 측정자료와 주관적 평가자료를 이용하여 회귀분석을 실시하였다. 기초기술 8개 측정항목의 결과에 대한 각 집단별 비교에서는 대학생이 현저하게 우수한 기록을 보였으며, 그 다음으로 고교생, 중학생 집단 순으로 나타났다. 그러나 각 측정항목에 대한 지도자의 주관적 평가점수에서는 대학생과 고교생 집단은 뚜렷한 차이가 보이지 않았다. 측정항목에 대한 측정자료와 각 항목에 대한 지도자의 주관적 평가자료의 관련성에 대한 상관분석 결과에 의하면 두 자료의 상관계수는 매우 낮았으며 전반적으로 객관적 측정자료와 지도자의 주관적 평가자료의 일치성은 매우 낮았다. 객관적 측정자료와 지도자의 주관적 평가자료의 관련성이 매우 낮게 나타난 결과에 대하여 다양한 관점에서 접근하여 차이점을 분석하였으며 전반적으로 스포츠 기술의 객관적 측정 및 평가의 타당성을 높이기 위한 방안을 논의하였다. 지도자의 종합경기력 평가에 영향을 미치는 요인을 알아보기 위해 실시한 회귀분석결과, 중학생 집단은 기술의 정확성과 기초체력에 관련된 변인, 즉 ``패스의 정확성``, ``돌파력``, 그리고 ``10m 왕복 드리블``이 가장 관련이 많았으며, 고교와 대학 집단은 전반적으로 복합적인 요인, 즉 지도자의 주관적 평가 항목인 ``경기감각``, ``연결동작``, ``킥 및 득점력``이 종합경기력 점수와 가장 관련이 높았다. The primary purpose of this study was to compare and analyze the characteristics between objective and subjective evaluations for soccer skill performances. Eight test items, such as heading, zigzag dribble, long kick, dribble shoot, shoot, centering, 30m dribble, and 10m shuttle dribble, were devised based on valid measurement procedures to investigate the ability of basic soccer skills. Data were collected from the 359 varsity soccer players (112 junior high, 133 senior high, and 114 university athletes) and analyzed using analysis of variance to compare the mean performance of each skill test among the 3 groups. Correlational methods were applied to investigate the relationship between the objective measures of the tests and the subjective measures of the corresponding skills evaluated by the coaches of the participating players. From these analyses the following conclusions were obtained. 1. As being expected, the comparison of the mean performances among the 3 groups revealed that the university group was prominent in all 8 skill tests, followed by the senior high and junior high school players. 2. It was shown that the coach`s subjective evaluation of the players` performance on each test was not quite in agreement with the field test scores. In most test items, there was a moderate size of correlation between the test scores and the coach`s evaluation. However, although it was statistically significant, the magnitude of correlation was not large enough to grant practically meaningful interpretations. 3. The results of multiple regression analyses indicated that the variables significantly contributing to the prediction of the overall playing ability consisted of mostly the subjective measures of the skills evaluated by the coaches.

체육측정평가학 : 혈액 분석기기의 신뢰도 및 타당도 분석

엄한주(HanJooEom),강성기(SeoungKiKang) 한국체육학회 2004 한국체육학회지 Vol.43 No.6

This study investigates the reliability and validity of blood lipid analyzers. Four blood analyzers, two different types (DT 60, Hitachi 7600-110,) with 2 instruments for each type, were used to compute the values of TC, HDL-C, and TG. Also, Lactic acid concentrations were repeatedly obtained by two instruments (DT 60 and Cobas integra, each type having 2 instruments). In all analyses, the same blood samples from each of 6 subjects were used to examine the repeatability and reproducibility of the values measured by all instruments. Data were analyzed to compute coefficient of variations (CV), variance components, correlations and various descriptive statistics with SPSS 11.0. HDL-C and TC values obtained by the same type of analyser for the 3 repeated samples from the same person under each condition were highly consistent, the CV for each being within 5%. However, when the values obtained from two different types of the analyzers were combined, the CV increased up to 12.3% for HDL-C and 21.7 ~ 24.5% for TC, indicating very low reproducibility of the analyzers. The correlations between the values of each measured variables with the same type of analyzers were very high, reaching above .97, but those values obtained by different types of analyzers were considerably lower. The analyses of variance component also revealed that the main source of error variances were mostly due to the difference in the types of analyzer. Thus, these results suggest that the different types of analyzers do not produce the same results although the 2 types are designed and manufactured to have the same function in analyzing blood lipid. For the TG and Lactic acid variables, the results indicated that there was insufficient evidence of both repeatability and reproducibility of the analyzers as the values obtained repeatedly from the same type of analyzers as well as those obtained from different types of analyzers were inconsistent. In general, the blood analyzers for TC and HDL-C were reliable, but not valid. However, the values of TG and Lactic acid were neither reliable nor valid as the values were varied considerably both within the same type and between the different types of analyzers. HDL-C and TC concentration is within coefficient of variation(CV) 5%. A similar Hitachi 7600-110 and DT 60 blood analyzer whole mean concentration of participants lower. However, defined as the calculated HDL-C (CV:12.8-16.36%), TC (CV: 21.68-24.47%), TC concentration should alter with change the analyzer to estimating blood range. Both HDL-C and TC concentration is above correlate .94 respectively. Therefor, to analyse HDL-C and TC, we found high mean different. Variance component estimation of HDL-C and TC are HDL-C(33.88%), TC(11.65%) which is different participant and another variance component estimation is different blood analyzer(HDL-C 6205%, TC 86.11%). To analyse TG and Lactic acid, we found estimation values of blood analyzer. Variance component estimation of TG and Lactic acid are TG(627%), Lactic add(87.9%) of different blood analyzer.

신뢰계수의 종류와 추정법에 관한 고찰

엄한주(Eom Han Joo) 한국체육측정평가학회 1999 한국체육측정평가학회지 Vol.1 No.1

반복측정자료의 구형성 조건과 분석결과의 이해

엄한주(Han Joo Eom) 한국체육측정평가학회 2007 한국체육측정평가학회지 Vol.9 No.2

ANOVA 의 가정 위반이 일반화가능도 계수의 추정에 미치는 영향 : 이원국면 교차설계

엄한주(Han Joo Eom) 한국체육측정평가학회 2000 한국체육측정평가학회지 Vol.2 No.1