Dissertations / Theses on the topic 'Kappa coefficient'

Quand on veut tester l'efficacite d'un traitement et que la reponse a ce traitement se presente sous la forme d'une duree de vie (eventuellement censuree), on a le plus souvent recours a des tests non-parametriques tels que les tests du logrank pondere. L'utilisation de tels tests repose sur l'hypothese que le traitement a ete attribue au hasard, c'est-a-dire que tous les individus avaient, a priori, la meme probabilite de recevoir le traitement. On propose une methode permettant de repondre a la question : quelle confiance accorder aux resultats d'une telle etude si l'hypothese d'equiprobabilite n'est pas verifiee ? l'objet de cette these est d'etudier la sensibilite des tests du logrank pondere a une heterogeneite qui remet en cause l'hypothese d'equiprobabilite dans l'attribution des traitements. On montre que, si le facteur d'heterogeneite est inconnu, il est possible de donner un encadrement de la loi de ces statistiques de tests. Les bornes de cet encadrement sont donnees d'une part par la configuration du facteur d'heterogeneite la plus favorable au traitement, et d'autre part par la moins favorable. Si on suppose que le traitement est attribue de facon independante a chaque individu, ou si on travaille sur des donnees appariees, on peut montrer que, meme en presence d'heterogeneite, les statistiques sont asymptotiquement normales. Dans le cas ou le nombre d'individus traites est fixe a priori, on montre sur des exemples que l'approximation normale semble raisonnable mais on ne prouve pas la validite de cette approximation. Enfin on compare les sensibilites des differents tests du logrank pondere et on montre que le moins sensible est le test du logrank, le plus sensible celui de gehan et, de facon generale, qu'un test est d'autant plus sensible qu'il donne aux individus des poids contrastes.

The Kappa coefficient is widely used in assessing categorical agreement between two raters or two methods. It can also be extended to more than two raters (methods).  When using Kappa, the shortcomings of this coefficient should be not neglected.  Bias and prevalence effects lead to paradoxes of Kappa. These problems can be avoided by using some other indexes together, but the solutions of the Kappa problems are not satisfactory. This paper gives a critical survey concerning the Kappa coefficient and gives a real life example. A useful alternative statistical approach, the Rank-invariant method is also introduced, and applied to analyze the disagreement between two raters.

Cotton crop condition was assessed from an analysis of multispectral aerial video imagery. Visible-near infrared imagery of two cotton fields was collected towards the end of the 1990 crop. The digital analysis was based on image classification, and the accuracies were assessed using the Kappa coefficient of agreement. The earliest of three images proved to be best for distinguishing plant variety. Vegetation index images were better for estimating potential yield than the original multispectral image; so too were multi-channel images that were transformed using vegetation indices or principal component analysis. The seedbed preparation rig used, the nitrogen application rate and three plant varieties, a weed species and two cotton cultivars, could all be discriminated from the imagery. Accuracies were moderate for the discrimination of plant variety, tillage treatment and nitrogen treatment, and low for the estimation of potential yield.

L'objectif de cette thèse consiste à développer une architecture de
fusion d'informations basée sur la théorie floue pour la segmentation
d'une cible à partir de plusieurs sources d'images. Notre application
principale porte sur la segmentation des images IRM multispectrales. Nous proposons une approche de segmentation automatique basée sur la fusion des caractéristiques extraites de chaque source d'image. Ces caractéristiques sont modélisées par des fonctions d'appartenance, obtenues à partir de fonctions analytiques, qui prennent en compte des connaissances a priori sur la possibilité d'appartenance à une cible (tumeur ou tissus cérébraux) donnée par l'expert, et aussi la gradation d'intensité du signal de la cible.

La segmentation d'une cible consiste finalement à fusionner les
différents degrés d'appartenance de la cible. Une étape supplémentaire basée sur une croissance 3D des régions floues est proposée pour améliorer le résultat de la fusion. Pour évaluer les résultats de segmentation représentés par un ensemble flou, une extension du coefficient Kappa de Cohen, nommée " Kappa flou " est proposée, qui est une méthode d'évaluation globale sur la proportion d'agrément d'un classement flou.

Cette architecture développée est mise en œuvre pour la segmentation des tumeurs cérébrales à partir des images IRM qui comprennent pour l'instant les séquences de routine : T1, T2 et densité de protons. Les résultats sur sept patients atteints de tumeur montrent l'efficacité de notre système.

L'objectif de cette thèse consiste à développer une architecture de fusion d'informations basée sur la théorie floue pour la segmentation d'une cible à partir de plusieurs sources d'images. Notre application principale porte sur la segmentation des images IRM multispectrales. Nous proposons une approche de segmentation automatique basée sur la fusion des caractéristiques extraites de chaque source d'image. Ces caractéristiques sont modélisées par des fonctions d'appartenance, obtenues à partir de fonctions analytiques, qui prennent en compte des connaissances à priori sur la possibilité d'appartenance à une cible (tumeur ou tissus cérébraux) donnée par l'expert, et aussi la gradation d'intensité du signal de la cible. La segmentation d'une cible consiste finalement à combiner les différents degrés d'appartenance de la cible. Une étape supplémentaire basée sur une croissance 3D des régions floues est proposée pour améliorer le résultat de la combinaison. Pour évaluer les résultats de segmentation représentés par un ensemble flou, une extension du coefficient Kappa de Cohen, nommée "Kappa flou" est proposée, qui est une méthode d'évaluation globale sur la proportion d'agrément d'un classement flou. Cette architecture développée est mise en oeuvre pour la segmentation des tumeurs cérébrales à partir des images IRM qui comprennent pour l'instant les séquences de base : T1, T2 et densité de protons (DP). Les résultats sur sept patients atteints de tumeur montrent l'efficacité de notre système.

Made available in DSpace on 2014-12-17T15:55:04Z (GMT). No. of bitstreams: 1 GabriellaCMS_DISSERT.pdf: 3451606 bytes, checksum: 1ee33015b3633fb79771b6160bff6a6c (MD5) Previous issue date: 2013-02-25
Coordena??o de Aperfei?oamento de Pessoal de N?vel Superior
The municipality of Areia Branca is within the mesoregion of West Potiguar and within the microregion of Mossor?, covering an area of 357,58 km2. Covering an area of weakness in terms of environmental, housing, together with the municipality of Grossos-RN, the estuary of River Apodi-Mossor?. The municipality of Areia Branca has historically suffered from a lack of planning regarding the use and occupation of land as some economic activities, attracted by the extremely favorable natural conditions, have exploited their natural resources improperly. The aim of this study is to quantify and analyze the environmental degradation in the municipality. Thus initially was performed a characterization of land use using remote sensing, geoprocessing and geographic information system GIS in order to generate data and information on the municipal scale, which may serve as input to the environmental planning and land use planning in the region. From this perspective, were used a Landsat 5 image TM sensor for the year 2010. In the processing of this image was used SPRING 5.2 and applied a supervised classification using the classifier regions, which was employed Bhattacharya Distance method with a threshold at 30%. Thus was obtained the land use map that was analyzed the spatial distribution of different types of the use that is occurring in the city, identifying areas that are being used incorrectly and the main types of environmental degradation. And further, were applied the methodology proposed by Beltrame (1994), Physical Diagnosis Conservationist under some adaptations for quantifying the level of degradation or conservation study area. As results, the indexes were obtained for the parameters in the proposed methodology, allowing quantitatively analyze the degradation potential of each sector. From this perspective, considering a scale of 0 to 100, sector A and sector B had value 31.20 units of risk of physical deterioration. And the C sector, has shown its value - 34.64 units degradation risk and should be considered a priority in relation to the achievement of conservation actions
O munic?pio de Areia Branca-RN est? inserido na mesorregi?o Oeste Potiguar e na microrregi?o de Mossor?, abrangendo uma ?rea de 357,58 km2. Compreende uma ?rea de fragilidade do ponto de vista ambiental, pois abriga, juntamente com o munic?pio de Grossos- RN, o estu?rio do rio Apodi-Mossor?. O munic?pio de Areia Branca vem sofrendo historicamente com a falta de planejamento no tocante ao uso e ocupa??o do solo, uma vez que algumas atividades econ?micas, atra?das pelas condi??es naturais favor?veis, t?m explorado os recursos naturais de forma inadequada. O objetivo deste estudo ? quantificar e analisar a degrada??o ambiental no referido munic?pio. Para isso, inicialmente foi realizada uma caracteriza??o do uso do solo, utilizando sensoriamento remoto, geoprocessamento e um sistema de informa??es geogr?ficas - SIG, visando gerar dados e informa??es na escala municipal, que possam servir de subs?dio para o planejamento ambiental e o ordenamento territorial da regi?o. Nessa perspectiva, utilizou-se uma imagem Landsat 5, sensor TM referente ao ano de 2010. No processamento desta imagem foi utilizado o SPRING 5.2 e aplicado uma classifica??o supervisionada atrav?s do classificador por regi?es, onde foi empregado o m?todo Bhattacharya Distance com um limiar 30%. Com isso foi obtido o mapa de uso do solo a partir do qual analisou-se a distribui??o espacial dos diferentes tipos de uso que ocorrem no munic?pio, identificando ?reas que est?o sendo utilizadas de maneira incorreta e os principais tipos de degrada??o ambiental. Em prosseguimento, aplicou-se a metodologia proposta por Beltrame (1994), o Diagn?stico F?sico-Conservacionista, sob algumas adapta??es, para obter a quantifica??o do n?vel de degrada??o ou conserva??o da ?rea de estudo. Como resultados, foram obtidos os ?ndices para os par?metros propostos na metodologia, permitindo analisar quantitativamente o potencial de degrada??o de cada setor. Nessa perspectiva, considerando uma escala de 0 a 100, o setor A e o setor B apresentaram valor 31,20 unidades de risco de degrada??o f?sica. E o setor C, demonstrou valor 34,64 unidades de risco de degrada??o, devendo ser considerado prioridade no tocante ? realiza??o de a??es conservacionistas

Made available in DSpace on 2017-05-12T14:47:32Z (GMT). No. of bitstreams: 1 Vanterlei _Artur Bier.pdf: 3411579 bytes, checksum: 9d27d1df8109289eb3cf3192d069b831 (MD5) Previous issue date: 2015-02-23
The precision agriculture (PA) is defined as the use of site specific management techniques that allow cropping management according to its needs and soil, in order to reduce impacts on the environment. The models valuation that interpolates field data and generates thematic maps is a task that requires in-depth knowledge on this issue. The Cohen Kappa index (K) is the most widely statistics used to compare thematic maps. The accuracy with which the spatial distribution maps of soil attributes are produced influences the implementation and use of PA. However its use has the disadvantage of providing variation in accordance with the use of different numbers of classes, adopted during the map generation process. Thus, this work aimed at selecting the best method among four interpolation ones (inverse distance, inverse distance squared, ordinary kriging and cokriging) using the index of selection of interpolators (ISI), proposed here, based on the contents of clay, copper and manganese, area elevation data and the apparent soil electrical conductivity. The selection among mathematical models and geostatistical interpolation was simplified using the ISI. The study also evaluated the influence in agreement K and Tau (T) indices when varying the number of confusion matrix classes in results that come from a 15.5 ha area, with typical Red Dystrophic soil in Céu Azul countryside, Paraná, Brazil, where soil properties were interpolated with the inverse distance, inverse square of distance, ordinary kriging and cokriging. According to this trial, K and T indices have been confirmed and varied widely of agreement for different numbers of classes. Thus, in order to solve this situation, it was developed K and T equivalent indices to compare thematic maps of quantitative data, using the relative deviation coefficient ,the absolute averaged deviation of interpolated data, the average, and standard deviation of the attribute original data. The result proved to be a good alternative to K and T indices based on the error matrix since it is independent of the classes number and shows a simpler calculation. The methodology was more efficient for situations when more than four classes are used.
A agricultura de precisão (AP) é definida como a utilização de técnicas que permitem manejo localizado de cultivo de acordo com as necessidades da cultura e do solo, para diminuir impactos no meio ambiente. A avaliação de modelos que interpolam dados de campo e geram mapas temáticos é uma tarefa que exige conhecimentos aprofundados no assunto. A precisão com que os mapas de distribuição espacial de atributos do solo são produzidos influencia a aplicação e a utilização da AP. O índice de concordância Kappa de Cohen (K) é a estatística mais utilizada em comparação de mapas temáticos. Entretanto, seu uso com dados quantitativos apresenta o inconveniente de proporcionar variação de concordância com a utilização de diferentes números de classes, adotada durante o processo de geração do mapa. Assim, o objetivo deste trabalho foi selecionar o melhor entre quatro métodos de interpolação (inverso da distância, inverso da distância ao quadrado, krigagem ordinária e cokrigagem) utilizando o índice de seleção de interpoladores (ISI), aqui proposto, a partir de teores de argila, cobre e manganês, dados de elevação do terreno e da condutividade elétrica aparente do solo. Com o uso do ISI, a seleção entre modelos determinísticos e estocásticos de interpolação ficou simplificada. O estudo avaliou ainda a influência nos índices de concordância K e Tau (T), quando se varia o número de classes da matriz de confusão, em resultados oriundos de uma área de 15,5 ha, com solo Latossolo Vermelho Distroférrico típico, no município de Céu Azul, Paraná, Brasil. A partir deste trabalho, foi confirmado que os índices Kappa e Tau apresentam grandes variações de concordância para diferentes números de classes. Para resolver esta situação desenvolveram-se os índices K e T alternativos para comparação de mapas temáticos de dados quantitativos, utilizando-se o coeficiente de desvio relativo, o desvio absoluto médio dos dados interpolados, a média e o desvio padrão dos dados originais do atributo. O resultado mostrou-se como boa alternativa aos índices de K e T baseados na matriz de erro por ser independente do número de classes e ser de cálculo mais simples. A metodologia mostrou-se mais eficiente para situações em que se utilizam mais que quatro classes

碩士
國立中央大學
統計研究所
106
In this paper, we construct a robust likelihood function for the agreement kappa/weighted kappa coefficient for clustered paired data in the case of three-category diagnostic outcome scenario. Utilizing this robust likelihood function, one can construct robust likelihood ratio (LR) statistic and LR-based confidence intervals without specifically modeling the intra-cluster correlation. We also make comparison between our robust likelihood approach and the nonparametric inferential method for kappa with paired data proposed by Yang and Zhou (2014, 2015) via simulations and real data analysis.

The study of interrater agreement and itnerrater reliability attract extensive attention, due to the fact that the judgments from multiple raters are subjective and may vary individually. To evaluate interrater agreement and interrater reliability, five different methods or indices are proposed: percentage of agreement, kappa coefficient, the correlation coefficient, intraclass correlation coefficient (ICC), and generalizability (G) theory. In this study, we introduce and discuss the advantages and disadvantages of these methods to evaluate interrater agreement and reliability. Then we review and explore the rank across these five indices by use of frequency in practice in the past five years. Finally, we illustrate how to use these five methods under different circumstances and provide SPSS and SAS code to analyze interrater agreement and reliability. We apply the methods above to analyze the data from Parent-Child Interaction System of global ratings (PARCHISY), and conclude as follows: (1) ICC is the most often used method to evaluate interrater reliability in recent five years, while generalizability theory is the least often used method. The G coefficients provide similar interrater reliability with weighted kappa and ICC on most items, based on the criteria. (2) When the reliability is high itself, different methods provide consistent indication on interrater reliability based on different criteria. If the reliability is not consistent among different methods, both ICC and G coefficient will provide better interrater reliability based on the criteria, and they also provide consistent results.

Rare-event phenomena are ubiquitous in nature. We propose a new strategy, kappa-dynamics, to model rare event dynamics. In this methodology we only assume that the important rare-event dynamics obey first-order kinetics. Exact rates are not required in the calculation and the reaction path is determined on the fly. kappa-dynamics is highly parallelizable and can be implemented on computer clusters and distributed machines. Theoretical derivations and several examples of atomic scale dynamics are presented. With single-molecule (SM) techniques, the individual molecular process can be resolved without being averaged over the ensemble. However, factors such as apparatus stability, background level, and data quality will limit the amount of information being collected. We found that the correlation function calculated from the finite-size SM rotational diffusion trajectory will deviate from its true value. Therefore, care must be taken not to interpret the difference as the evidence of new dynamics occurred in the system. We also proposed an algorithm of single fluorophore orientation reconstruction which converts three measured intensities {I₀,I₄₅,I₉₀} to the dipole orientation. Fluctuations in the detected signals caused by the shot noise result in a different prediction from the true orientation. This difference should not be interpreted as the evidence of the nonisotropic rotational motion. Catalytic reactions are also governed by the rare-events. Studying the dynamics of catalytic processes is an important subject since the more we learn, the more we can improve current catalysts. Fuel cells have become a promising energy source in the past decade. The key to make a high performance cell while keeping the price low is the choice of a suitable catalyst at the electrodes. Density functional theory calculations are carried out to study the role of geometric relaxation in the oxygen-reduction reaction for nanoparticle of various transition metals. Our calculations of Pt nanoparticles show that the structural deformation induced by atomic oxygen binding can energetically stabilize the oxidized states and thus reduces the catalytic activity. The catalytic performance can be improved by making alloys with less deformable metals.
text

碩士
國立中央大學
統計研究所
107
The kappa coefficient is an index that is widely used to measure the agreement between different raters or different methods. In this paper, we construct a robust likelihood function for the quality of two kappa coefficients for correlated paired data in the case of binary diagnostic outcome scenario. Using this robust likelihood function, we can make correct statistical inferences without modeling the correlated joint distribution in paired data. We provide simulations and real data analyses to demonstrate the robust likelihood approach.

碩士
國立彰化師範大學
統計資訊研究所
97
Many studies in medical, industry and business fields commonly use questionnaires and diagnoses reported by different raters or instruments to evaluate the agreement of multiple ratings. The most popular indices of agreement for a categorical response are the kappa measures, Cohen’s kappa for binary outcomes and weighted kappa for ordinal outcomes (Cohen 1960, 1968). However, when raters assess the same observation on two or more occasions, these ratings are dependent, and, therefore one must take into account the correlation between the kappa estimates for making inference. In this thesis, we focus on estimating kappa measures on inter and intra-agreement using the weighted least squares (WLS) approach for dependent categorical data and testing the equality of the dependent kappa coefficients by using PROC CATMOD in SAS. We modified the SAS code in Barnhart and Williamson (2002) to analyze dependent agreement data with more than two raters and to use different numbers of categories. In addition, it will be compared with the results using generalized estimating equation (GEE) for Cohen’s kappa and weighted kappa. Simulation studies are conducted to evaluate the performance under WLS and GEE approaches. Four applications of repeated ratings by multiple raters are illustrated, including similarity in myopic status for twin pairs with respect to zygosity, conformity between an optometric device and ophthomologist in diagnoses of high myopia, test-retest radiographic examinations by seven endodontists, and comparison between a computerized planimetry method and direct visual assessment for assessing cervical ectopy. The results reveal that the performance of WLS is better than that of GEE.

Introduction Colour codes are used extensively in railways to convey specific information governing movement of trains and equipment on the track. One such task is the railway traffic control display that uses colour coded video display terminals (VDTs) to convey information of the signal status, train movements and track status to the railway dispatcher. Because individuals with colour vision deficiencies (colour-defectives) may have problems with these colour-related tasks, questions were raised about the suitability of colour vision defectives to work as railway dispatchers. In order to answer that, a VDT based Dispatch Colour Vision Test based on the actual railway traffic display was developed previously. Purpose The main purpose of this thesis is to establish the pass/fail scores and repeatability of the VDT based Dispatch Colour Vision Test that resulted from the previous work. Secondly, the study will also examine whether clinical colour vision tests can predict the performance on the practical task. Methods The Dispatch colour vision test was divided into three parts based on the colour sets that the dispatcher had to recognize. The testing computer system used the the same RGB colour settings, graphics card and monitor as in railway dispatch centres. Subjects viewed the display colours and entered their responses by using a mouse. One hundred colour-normals and fifty two colour-defectives participated in the initial session. The test was repeated approximately after 10 days. Ninety three colour-normals (93%) and 44 (85%) colour-defectives participated in the second session. The total number of errors and time to complete the test was recorded. Results Pass/Fail on the VDT Dispatch colour vision test was based on colour-normal errors. Ignoring orange-red errors, two errors were allowed in the first session and one error was allowed in the second session. Based on this criterion, 42% of colour vision defectives could perform as well as colour normal subjects. The kappa coefficient of agreement between the sessions for the colour-defectives was 0.85. Detailed analysis between the colour differences and the errors showed only a weak correlation between the two. However, the general trend was that colour-defectives made more errors on colours that were near or along the same lines of confusions and the colours were nearly equal in luminance. Nevertheless, the interaction between luminance and location with respect to the lines of confusion was not easy to interpret. The time to complete the task for the colour-defectives who passed the test took 14% longer than colour-normals and colour-defectives who failed took 30% longer than colour-normals. All groups showed a similar learning effect with an 18% reduction in mean times to complete the task at the second session. There was no significant correlation between the number of errors and time to complete or the clinical tests and completion times for any of the groups. Clinical colour vision tests have limited value in predicting performance of colour-defectives on the Dispatch test. Logistic analysis results showed that the Farnsworth D-15 along with the Nagel was the best predictor of the VDT Dispatch colour test pass/fail results. However, these results were similar to using the Farnsworth D-15 test alone. Ninety-five percent of the individuals who failed the Farnsworth D-15 also failed the Dispatch test. However, approximately 25% of the individuals who passed the Farnsworth D-15 failed the VDT Dispatch colour test which is an unacceptable false negative rate. These results indicate the Farnsworth D-15 can only be used to predict who is likely to fail the dispatch test. Conclusions Forty two percent of colour vision defectives could perform as well as colour-normals in identifying VDT railway display colours and time to complete the task. Clinical colour vision tests were inadequate predictors of performance in practical task, overall. However, the Farnsworth D-15 was a very good predictor of who would fail the VDT Dispatch test. Hence a practical VDT Dispatch test may be needed to test individuals who would want to work as railway dispatchers.

This research work concentrate on developing digital soil maps to support field based plant phenotyping research. We have developed soil organic matter content (OM), cation exchange capacity (CEC), natural soil drainage class, and tile drainage line maps using topographic indices and aerial imagery. Various prediction models (universal kriging, cubist, random forest, C5.0, artificial neural network, and multinomial logistic regression) were used to estimate the soil properties of interest.