Dissertations / Theses on the topic 'Coefficient of concordance'

La fiabilité d'une mesure fait référence à sa reproductibilité lorsqu'elle est répétée de manière aléatoire sur le même sujet et constitue une propriété métrologique clé pour toutes les mesures. Le coefficient de corrélation intra-classe de concordance à deux voies (ICCa) est un paramètre statistique utilisé pour quantifier la fiabilité inter-observateurs des mesures continues (ou qualitatives ordinales). Il constitue un indice de fiabilité central recommandé par les agences réglementaires. Néanmoins, ses estimateurs sont biaisés et de nombreuses solutions ont été essayées pour améliorer son intervalle de confiance (IC). Les travaux les plus récents indiquent qu'aucune méthode ne fonctionne bien en cas de non-normalité des données, difficilement détectable à partir des observations, et lorsque le nombre de sujets ou d'évaluateurs est limité, ce qui est plutôt le cas en pratique. De plus, aucune transformation de stabilisation de la variance (VST) ni aucun test de comparaison statistique ne sont disponibles pour l'ICCa. L'objectif de cette thèse est donc de contribuer au développement de méthodes palliant le manque d'outils inférentiels pour l'ICCa. Dans un premier temps, nous proposons de nouvelles méthodes asymptotiques pour l'intervalle de confiance de l'ICCa, le calcul de la taille de l'échantillon des sujets et des évaluateurs, et un test de rapport de vraisemblance pour comparer deux ICCa. Ensuite, dans un deuxième travail, nous développons trois VST, améliorant les propriétés de l'intervalle de confiance pour les études de fiabilité inter-évaluateurs de taille modérée, et la synthèse de plusieurs ICCa dans le contexte de méta-analyses. Enfin, dans un troisième travail, des méthodes de rééchantillonnage spécifiques sont proposées, en combinaison avec la meilleure VST, pour améliorer les performances de l'intervalle de confiance de l'ICCa sans l'hypothèse de normalité et avec de petits échantillons. Il s'agit d'un travail de méthodologie biostatistique, avec des évaluations par simulation des méthodes introduites, et des applications à plusieurs jeux de données réelles issues d'études de fiabilité inter-évaluateurs et de méta-analyses
Measurement's reliability refers to its reproducibility when it is randomly repeated on the same subject and is a key metrological property for each measurement validation. The two-way intra-class correlation coefficient of agreement (ICCa) is a statistical parameter used to quantify the inter-rater reliability of continuous (or ordinal qualitative) measurements. It constitute a central reliability index recommended by the regulatory agencies. Nevertheless, its estimators are biased and a lot of solutions have been tried facing to its confidence interval (CI) problem. The latest works indicate that no method works well with a hard-to-detect violation of normality and when the number of subjects OR raters is limited, which is rather the case in practice. Furthermore, no variance stabilizing transformation (VST) nor statistical comparative test are available for the ICCa. The aim of this thesis is therefore to contribute to the development of methods that remedy the lack of the inferential tools for the ICCa. At a first step, we propose new asymptotic methods for the ICCa confidence interval, the calculation of the needed sample size of subjects and raters, and a likelihood ratio test to compare two ICCa. Then, in a second work, we develop three VSTs, improving the properties of the confidence interval for inter-rater reliability studies of moderate sample size, and the synthesis of several ICCa in the context of meta-analyses. Finally, in a third work, dedicated resampling methods are proposed, in combination with the best VST, to improve the ICCa confidence interval performances in case of non-normality with small sample size. It is above all a work of biostatistical methodology, with simulation evaluations of the introduced methods, and applications to several real data sets from inter-rater reliability studies and meta-analyses

Thesis (MBA.) -- University of Limpopo, 2016
Project systems allow project managers to carry out work in a professional and well organised manner. These systems are created and maintained to advance project performance. Eskom spends a lot of resources on the creation and maintenance of project systems. The literature on project systems shows that, in spite of advancement in project managements processes, systems and tools, project success has not significantly improved. This problem raises questions about the value and effectiveness of project systems. Therefore this paper reports about the correlation between the degree of project manager’s compliance to project systems and project performance in Eskom Distribution Limpopo Operating Unit. The study looked at the performance of 10 projects and used empirical data on designers, planners, managers and project managers working in Eskom Distribution LOU to measure the compliance level of employees to project systems when carrying out the 10 chosen projects or any other projects not listed. A total of 45 completed questionnaires were analysed. Correlation analysis tests found a negative correlation between project manager’s project systems compliance level and project performance in terms of schedule and cost. The conclusion found was that as the compliance level on project systems increases, project performance decreases. Meaning there is an inversely proportional relationship between project system compliance level and project performance. Additionally, a lower level of knowledge than expected on the project managers, designers, and planners was found. Keywords: Project performance, Project systems, Correlation, Adherence,

Es desenvolupa una metodologia que combina algoritmes exploratoris i iteratius que pren com a base el concepte de concordança del coeficient de correlació W de Kendall. A partir de conjunts complexos d'ítems i les seves puntuacions (activitats d'avaluació, problemes, pràctiques, indicadors de rendiment, mesuraments i, en general, qualsevol variable quantitativa) s'aconsegueixen identificar conjunts concordants, és a dir, estructures explicatives de la dificultat del conjunt de ítems. La lectura d'aquests conjunts es pot fer de dues maneres gràcies a la informació que proporcionen els algoritmes creats: a) Transversal, de manera que els conjunts concordants formen el "esquelet de dificultat" del problema i són, per tant, referències per estudiar la resta de puntuacions en altres ítems. b) Longitudinal, mitjançant l'estudi del que anomenem Mapes de Concordança, els quals mostren l'evolució dels conjunts concordants en anar incorporant elements en cada iteració, i detectant per tant els ítems discordants respecte a conjunts estables donats. L'aplicació d'aquesta metodologia és molt àmplia, i és susceptible de ser posada en pràctica en qualsevol camp de les ciències on es realitzin mesures i es vulgui observar les variacions creuades entre individus. En aquesta investigació s'aplica a l'assignatura de Laboratori de mesures electròniques. Els resultats aporten llum a la manera de millorar les pràctiques existents i de construir-ne de noves en el futur.
Se desarrolla una metodología que combina algoritmos exploratorios e iterativos que toma como base el concepto de concordancia del coeficiente de correlación W de Kendall. A partir de conjuntos complejos de ítems y sus puntuaciones (actividades de evaluación, problemas, prácticas, indicadores de rendimiento, mediciones y, en general, cualquier variable cuantitativa) se consiguen identificar conjuntos concordantes, es decir, estructuras explicativas de la dificultad del conjunto de ítems. La lectura de estos conjuntos puede hacerse de dos maneras gracias a la información que proporcionan los algoritmos creados: a) Transversal, de manera que los conjuntos concordantes forman el “esqueleto de dificultad” del problema y son, por tanto, referencias para estudiar el resto de puntuaciones en otros ítems. b) Longitudinal, mediante el estudio de lo que denominamos Mapas de Concordancia, los cuales muestran la evolución de los conjuntos concordantes al ir incorporando ítems en cada iteración, y detectando por tanto los ítems discordantes respecto a conjuntos estables dados. La aplicación de dicha metodología es muy amplia, siendo susceptible de ser puesta en práctica en cualquier campo de las ciencias donde se realicen mediciones y se quiera observar las variaciones cruzadas entre individuos. En la presente investigación se aplica a la asignatura de Laboratorio de medidas electrónicas. Los resultados arrojan luz sobre la manera de mejorar las prácticas existentes y de construir otras nuevas en el futuro.
We have developed a new methodology that combines both exploratory and iterative algorithms which is based on the concept of concordance of the correlation coefficient W of Kendall. Starting from complex sets of items and their scores (assessment activities, problems, practices, performance indicators, measurements and, in general, any quantitative variable) it is possible to identify consistent sets, i.e., structures that explain the difficulty of the set of items. Reading these sets can be done in two different ways thanks to the information provided by the algorithms created: a) Transversal, so that the concordant sets form the "backbone difficulty" of the problem and, therefore, they become references to study the remaining scores on other items. b) Longitudinal, by studying what we call concordance maps, which show the evolution of the consistent sets when new items are incorporated after each iteration and thus detecting the discordant items with regard to given stable sets. The application of this methodology is very broad, being useful in any field of science where measurements are carried out and you may want to see cross-variations among individuals. In this research it has been applied to the subject Electronic Measurements Laboratory. The results point to different ways to improve the existing practices and create new ones in the future.

碩士
國立彰化師範大學
統計資訊研究所
102
In clinical studies, we concern with assessing agreement between different observers or methods. The concordance correlation coefficient (CCC) and intraclass correlation coefficient (ICC) are the two popular indices for measuring agreement on continuous data. In linear mixed models, the equivalence between ICC and CCC can be established through variance components (VC). In this thesis, we focus on estimating CCC using VC and U-statistics approaches. Simulation studies are conducted to compare the performance of VC and U-statistics approaches with and without explanatory variables. An application of conformity between two optometric devices is used for illustration.

З використанням кореляційного та генеалогічного аналізу оцінено рівень подібності дітей і батьків за типом темпераменту. У асоціації темпераменту у батьків та дітей виявлений материнський ефект.
The level of similarity between children and parents by their type of temperament is rated using the correlation and genealogical analysis. In the association of temperament in parents and children the maternal effect is revealed.

碩士
國立彰化師範大學
統計資訊研究所
106
The concordance correlation coefficient (CCC) is an indice that can measure the agreement between observers. In the past, the classical concordance correlation coefficient is applied to measure the agreement between two observers with continuous data. However, it is common that we need to assess the agreement among multiple observers for discrete responses in clinical research. In this paper, we introduce how to use a variance components (VC) method to assess the agreement among multiple observers for countable replicated readings. We assume that under a three-way generalized linear mixed model, the subject effect, the subject-by-observer interaction effect and the subject-by-time interaction effect are assumed as random effects, while the observer effect, the time effect and the observer-by-time interaction are assumed to be fixed. In the simulation studies, we use the VC approach to assess the intra- and inter- and total-CCCs and compare the performance of VC and U-Statistics (US) methods. It is concluded that no matter what sample size is used, the mean square errors of VC are smaller than those of US, and the 95% coverage rates of VC remain at the nominal coverage. Finally, an application of reproducibility of corticospinal diffusion tensor tractography study is illustrated.

碩士
國立彰化師範大學
統計資訊研究所
105
Variance components (VC) is an approach for estimating concordance correlation coefficients (CCC) adjusting for covariates, and allowing dependency between replicated samples. However, under VC, a model including all potential explanatory variables may lead to biased parameter estimates. To overcome this problem, the estimation of CCC using the VC approach, as well as applying the conditional Akaike information criterion (CAIC) and conditional Bayesian information criterion (CBIC) for model selection is applied. Simulation studies are conducted to compare linear mixed models (LMM), with and without model-selection via CAIC and CBIC, and we use three indices of intra-, inter- and total concordance correlation coefficients (CCCs) among multiple observers to evaluate agreement for continuous replicated readings. Two applications are illustrated: an assessment of conformity between three optometric devices and an evaluation of agreement in degree of myopia for monozygotic twins.

碩士
國立彰化師範大學
統計資訊研究所
106
In clinical medical researches, repeated measurements over time will be assessed on count data which is assumed to be distributed as Poisson or negative binomial distributions. In this thesis, we use generalized linear mixed models (GLMMs) for analyzing longitudinal Poisson data. In addition, we develop three indices of intra-, inter- and total concordance correlation coefficients (CCCs) among multiple observers to evaluate agreement for countable replicated readings by using a variance components (VC) approach to estimate CCCs. In simulation studies, we compare the performance of the VC and U-statistics approaches for repeated measurements from longitudinal Poisson data. An application of corticospinal diffusion tensor tractography study is used for illustration.

碩士
國立彰化師範大學
統計資訊研究所
107
In clinical studies, concordance correlation coefficient (CCC) is the most popular measures of agreement for a continuous scale. However, sample selection bias has long been recognized in many fields including clinical trials, epidemiology studies, genome-wide association studies, and wildlife management area. Therefore, the focus of this thesis is on evaluating how subject distribution and sample size affect the estimation of CCC under the existence of sampling bias. We use the variance components (VC) method to assess the intra-, inter- and total-CCCs and compare the performance of VC and U-Statistics (US) methods for linear mixed models (LMM). Simulation results show that CCC from the non-normal distribution is overestimated. Hence, we proposed a resampling procedure to reduces the influence of subject distribution on the estimation of CCC. Finally, an application of corticospinal diffusion tensor tractography study is used for illustration.

碩士
國立彰化師範大學
統計資訊研究所
107
In continuous clinical studies, concordance correlation coefficient (CCC) is widely used to assess agreement between two repeated measurements. It is also used to evaluate agreement of discrete data. However, CCC estimates may suffer from a serious bias caused by sampling bias. Therefore, the focus of this thesis is on proposing a variance components (VC) approach to assess intra-, inter- and total agreement among multiple observers under Poisson mixed-effects models for longitudinal Poisson data. In addition, in order to reduce effects of sampling bias, we develop a resampling approach to resample samples from non-Poisson distributions, and therefore the bias-corrected resamples will approximately follow a Poisson distribution. Simulation studies are conducted to compare the performance of the VC and U-statistics approaches with different sample sizes for repeated measurements from longitudinal non-Poisson data. In conclusion, the VC approach with the proposed resampling approach will reduce the influence of sampling bias on the estimation of CCC.

碩士
國立彰化師範大學
統計資訊研究所
104
In clinical studies, assessing agreement of multiple reading plays an important role in the evaluation of dependent continuous measurement scales. The replicated measurements may be produced by multiple observers or methods for each subject. In this paper, we propose three indices of intra-, inter- and total concordance correlation coefficients (CCCs) among multiple observers to evaluate agreement for continuous replicated reading. Furthermore, the three indices of agreement through variance components (VC) from a three-way linear mixed model (LMM) with consideration of the correlation structure of longitudinal repeated measurements are developed as well. Simulation studies are conducted to compare the performance of VC and U-statistics approaches. An application of conformity between three optometric devices is used for illustration.