本論文探討了以廣義線性混合模式(GLMM)結合時間及空間效果的時間空間模式，以將地區特性、人口特徵等變數，及時間變數納入模式中。有關時間效果可用B-Spline方法建構固定或隨機的時間趨勢平滑函數，而空間效果則是將各地區的隨機效果以條件自我相關模式(CAR)描述。實證部份則是應用GLMM模式分析台灣本島350個鄉鎮市區自民國八十八年到九十一年的肝癌就診資料，依性別、年齡層加以整理，並將年齡層分為0~19歲、20~39歲、40 ~59歲、60歲以上，分別代表少、青、壯、老等四個年齡層；再採用GLMMGibbs結合R軟體各個資料集分別配適時間空間模式，估計各地區之相對風險並繪製疾病地圖，據以找出各年估計的相對風險高的地區。

Phosphoglucosamine mutase (GlmM), an enzyme taking part in biosynthesis of cell wall, has been recently proven to be essential for Streptococcus pneumoniae. The main goal of this thesis was to prove in vivo that GlmM serine residues S99 and S101 phosphorylation is essential while the necessity of it was already proven indirectly based on transformation efficiency. For this purpose we have prepared a strain with two copies of the glmM gene - the first one with amino acid changes on monitored serine residues located at native locus; and the second ectopic copy of the wild allele of glmM gene under control of inducible zinc promoter. We have observed morphology, growth, and GlmM expression with and without the presence of an inductor. All the observed parameters show that the cells are not viable without ectopic glmM expression, thus the essential protein GlmM is functional only when phosphorylated on S99 and S101 residues. Further, we have attempted to localize the enzyme in the S. pneumoniae cell. We have fused GlmM with fluorescent marker GFP and by using the florescent microscopy we have proved that GlmM is cytoplasmic protein. Another goal of this thesis was to find an unknown third phosforylation site of the GlmM protein which is dependent on the protein kinase StkP. From in vitro kinase assay and...

?rea de concentra??o: Produ??o Animal. / Submitted by Rodrigo Martins Cruz (rodrigo.cruz@ufvjm.edu.br) on 2015-11-16T13:54:40Z No. of bitstreams: 2 license_rdf: 23898 bytes, checksum: e363e809996cf46ada20da1accfcd9c7 (MD5) diogo_anastacio_garcia.pdf: 241254 bytes, checksum: 3f5025c7cbe992519b20507c5901711c (MD5) / Approved for entry into archive by Rodrigo Martins Cruz (rodrigo.cruz@ufvjm.edu.br) on 2015-11-16T13:55:28Z (GMT) No. of bitstreams: 2 license_rdf: 23898 bytes, checksum: e363e809996cf46ada20da1accfcd9c7 (MD5) diogo_anastacio_garcia.pdf: 241254 bytes, checksum: 3f5025c7cbe992519b20507c5901711c (MD5) / Made available in DSpace on 2015-11-16T13:55:28Z (GMT). No. of bitstreams: 2 license_rdf: 23898 bytes, checksum: e363e809996cf46ada20da1accfcd9c7 (MD5) diogo_anastacio_garcia.pdf: 241254 bytes, checksum: 3f5025c7cbe992519b20507c5901711c (MD5) Previous issue date: 2010 / Funda??o de Amparo ? Pesquisa do estado de Minas Gerais (FAPEMIG) / Objetivou-se avaliar a aplica??o dos modelos lineares generalizados mistos na avalia??o gen?tica da prenhez precoce, comparando as fun??es de liga??o probit e logit, utilizando dados simulados e reais, al?m de avaliar os efeitos na variabilidade gen?tica e na sele??o de reprodutores quando diferentes defini??es desta caracter?stica s?o adotadas. O processo de simula??o foi determinado pela transforma??o aplicada para se obterem as probabilidades de prenhez das novilhas e da fun??o de liga??o utilizada na an?lise dos dados, sendo constru?das as estruturas de simula??o: logit-logit (LL), logit-probit (LP), probit-logit (PL) e probit-probit (PP). Foram adotadas distintas porcentagens de f?meas precoces (%FP): 5, 10, 15, 20, 25 e 30%. Para cada cen?rio de simula??o, constru?dos pela combina??o das estruturas e %FP, foram realizadas 100 repeti??es, sendo o valor param?trico da herdabilidade (h2) igual a 0,40. Nos dados reais, a prenhez precoce foi definida aos 15, 17, 19 e 21 meses, representadas por PP15, PP17, PP19 e PP21. A implementa??o da simula??o e a estima??o dos par?metros e predi??o dos valores gen?ticos foram realizadas no software R. No estudo de simula??o, as estimativas obtidas para a herdabilidade ( ) foram comparadas com o valor param?trico por meio do Erro Quadr?tico M?dio (EQM). Correla??es de Pearson (CP), entre os valores gen?ticos preditos e reais, e a porcentagem de touros em comum, entre a classifica??o real e predita, considerando apenas 10% dos touros com maiores valores gen?ticos (TOP10) foram calculadas. Para os dados reais, correla??es de Pearson, entre os valores gen?ticos preditos pelas fun??es de liga??o, e a TOP10, entre a classifica??o predita pela logit e probit e em cada fun??o entre PP15, PP17, PP19 e PP21 foram mensuradas. Al?m disso, para comparar o ajuste dos modelos, foram calculados os crit?rios de informa??o Bayesiano de Schwarz (BIC) e de Akaike (AIC). Considerando a simula??o, as estruturas LP e PL apresentaram resultados inferiores a LL e PP. Os valores de , EQM, correla??es de Pearson e TOP10, para as estruturas LL e PP, foram muito pr?ximos. Para os dados reais, as para PP15, PP17, PP19 e PP21, foram pr?ximas entre logit e probit, com exce??o da PP15. As CP e a TOP10 entre as fun??es foram altas. O AIC e BIC apresentaram-se semelhantes entre as fun??es, independente da classe de prenhez estudada. A TOP10, considerando a predi??o da mesma fun??o de liga??o, entre PP17-PP19, PP17-PP21 e PP19-PP21 foram de moderadas a alta. Entretanto, a TOP10 entre PP15-PP17, PP15-PP19 e PP15-PP21 podem ser consideradas baixas, independente da fun??o utilizada. ? necess?rio comparar o ajuste das fun??es de liga??o, uma vez que ao simular dados na escala logit e alis?-los por probit, e vice-versa, os resultados nas estimativas dos par?metros e predi??o n?o foram satisfat?rios como aqueles apresentados quando se gerou e analisou os dados por meio da mesma escala e fun??o de liga??o. Para as classes de precocidade estudadas, os modelos apresentaram estimativas de par?metros gen?ticos, predi??es dos efeitos aleat?rios e ajuste dos modelos muito semelhantes. A classifica??o dos touros foi consideravelmente diferente entre PP15-PP17, PP15-PP19 e PP15-PP21. / Disserta??o (Mestrado) ? Programa de P?s-Gradua??o em Zootecnia, Universidade Federal dos Vales do Jequitinhonha e Mucuri, 2010. / ABSTRACT This study aimed to evaluate the application of generalized linear mixed models on genetic evaluation of early pregnancy, by comparing probit and logit functions and using simulated and real data. Besides this, effects on genetic variability and on selection of sires under different definition are adopted. Simulation process was determined by the transformation applied to obtain probabilities of heifer pregnancy and the function used to analyze data, being simulation structures: logit-logit (LL), logit-probit (LP), probit-logit (PL) and probit-probit (PP). Different percentages of early female (%EF) were adopted: 5, 10, 15, 20, 25 and 30%. To each simulation, formed according to structures and %EP combination, there were 100 repetitions, being 0.40 the parametric value of heritability (h2). At real data, early pregnancy was determined at 15, 17, 19 and 21 months, represented by EP15, EP17, EP19 and EP21. To implement simulation, estimate parameters and predict genetic values, software R was used. In simulation study, estimative obtained to heritability ( ) were compared to parametric value through mean-square error (MSE). Pearson correlations (PC) were calculated between genetic real or predicted values and percentage of bulls in common, between real and predicted classification, considering only 10% of bulls with higher genetic values (TOP 10). To real data, there was Pearson Correlation between genetic values predicted by functions and the TOP10, between classification predicted by logit and probit and in each function between EP15, EP17, EP19 and EP21. In order to compare models adjustment, information criteria of Bayesiano of Schwarz (BIC) and Akaike (AIC) were calculated. Considering the simulation, LP and PL structures showed results inferior to LL and PP. Values of , MSEEQM, Pearson correlation and TOP10, to as LL and PP, were near. To real data, to EP15, EP17, EP19 and EP21, were near between logit and probit, except to EP15. AIC and BIC were similar among functions, independently of pregnancy class studied. Considering prediction of same function between EP17-EP19, EP17-EP21 and EP19-EP21 was from moderated to high. However, TOP10 between EP15-EP17, EP15-EP19 and EP15-EP21 can be considered low in any function. It is necessary to compare function adjustment because when data is simulated in logit and analyzed by probit, and vice-versa, results on estimated and predicted parameters were not satisfactory as the one shown when data were created and analyzed by the same scale and function. In precocity classes studied, the models showed estimative of genetic parameters, random effect prediction and models adjustment very similar. Bulls? classification was very different EP15-EP17, EP15-EP19 and EP15-EP21.

River flow regime is fundamental in determining lotic fish communities and populations, and especially of salmonid fishes. Quantifying the effects of human induced flow alteration on salmonids is a key question for conservation and water resources management. While qualitative responses to flow alteration are well characterised, a more intractable problem is quantifying responses in a way that is practical for environmental management. Using data drawn from the Environment Agency national database, I fitted generalised linear mixed models (GLMMs) using Bayesian inference to quantify the response of salmonid populations to the effects of impounding rivers, flow loss from rivers due to water abstraction, and the mitigating effects of flow restoration. I showed that in upland rivers downstream of impounded lakes, the magnitude of antecedent summer low flows had an important effect on the late summer abundance of 0+ salmonids Atlantic salmon (Salmo salar) and brown trout (Salmo trutta). In contrast, the abundance of 1+ salmon and brown trout appeared to be largely unresponsive to the same flows. I demonstrated that short-term flow cessation had a negative impact on the abundance of 1+ brown trout in the following spring, but that recovery was rapid with negligible longer-term consequences. I further established that flow restoration in upland streams impacted by water abstraction provided limited short-term benefits to salmonid abundance when compared with changes at control locations. However, while benefits to salmonid abundance were limited, I detected important benefits to the mean growth rates of 0+ and 1+ brown trout from flow restoration. I discuss the implications of my findings for salmonid management and conservation and propose a more evidence-based approach to fishery management based on robust quantitative evidence derived using appropriate statistical models. The current approach to flow management for salmonids requires revision and I recommend an alternative approach based on quantitative evidence.

Self-rated health (SRH) is an outcome commonly studied by demographers, epidemiologists, and sociologists of health, typically measured using an ordinal scale. SRH is analyzed in cross-sectional and longitudinal studies for both descriptive and inferential purposes, and has been shown to have significant validity with regard to predicting mortality. Despite the wide spread use of this measure, only limited attention is explicitly given to its unique attributes in the case of longitudinal studies. While self-rated health is assumed to represent a latent continuous and dynamic process, SRH is actually measured discretely and asymmetrically. Thus, the validity of methods ignoring the scale of measurement remains questionable. We compare three approaches to modeling SRH with repeated measures over time: linear multilevel models (MLM or LGM), including corrections for non-normality; and marginal and conditional ordered-logit models for longitudinal data. The models are compared using simulated data and illustrated with results from the Health and Retirement Study. We find that marginal and conditional models result in very different interpretations, but that conditional linear and non-linear models result in similar substantive conclusions, albeit with some loss of power in the linear case. In conclusion, we suggest guidelines for modeling self-rated health and similar ordinal outcomes in longitudinal studies. / text

Inference in generalized linear mixed models (GLMM) remains a topic of debate. Baayen, Davidson, and Bates (2008) outlines criticism against conventional ways of performing inference for GLMMs. There are various alternatives proposed but lit- tle consistency is found on which is the most reasonable. Our focus is on assessing temporal trends for mainly ecological count data. That is, we hope to provide a prag- matic approach to Poisson GLMMs for ecological researchers within the statistical programming environment R. To achieve this, we start by providing a description of the selected estimation and inferential procedures. We then complete a large scale simulation to evaluate each of the estimation methods. We implement a power analy- sis to assess each of the selected inferential procedures. We then go on to apply these procedures to data sampled by The National Parks of Canada. Finally, we conclude by giving a summary of our ?ndings and outlying work for the future.

The objective of this thesis is to examine random effect models applied to binary data. I will use classical and Bayesian inference to fit generalized linear mixed models to a specific data set. The data analyzed in this thesis comes from a study examining the injection practices of needle exchange clientele in Victoria, B.C. focusing on their risk networks. First, I will examine the application of social network analysis to the study of injection drug use, focusing on issues of gender, norms, and the problem of hidden populations. Next the focus will be on random effect models, where I will provide some background and a few examples pertaining to generalized linear mixed models (GLMMs). After GLMMs, I will discuss the nature of the injection drug use study and the data which will then be analyzed using a GLMM. Lastly, I will provide a discussion about my results of the GLMM analysis along with a summary of the injection practices of the needle exchange clientele. / Graduate / 0463

American black duck (Anas rubripes) populations declined throughout North America from 1950–1990, but the breeding population since has stabilized. However, limited information exists on black ducks in the Mississippi Flyway, where wintering populations continue to decline. I radiomarked 111 female black ducks at Tennessee National Wildlife Refuge (TNWR) in winters 2010–2012 to estimate winter survival and investigate patterns of habitat selection. Winter survival (83–85%) was greater than or comparable to previous estimates for black duck populations in North America. Interval survival increased 0.6% with a 100 g increase in body mass, but survival differed between years and waterfowl hunting seasons relative to body mass. Black ducks selected habitats on TNWR and emergent/scrub-shrub wetlands throughout winter regardless of hunting season or time of day. High winter survival rates and consistent use of TNWR suggest the refuge provides an important complex of habitats for black ducks wintering in Tennessee.

Une forte redondance des variables explicatives cause de gros problèmes d'identifiabilité et d'instabilité des coefficients dans les modèles de régression. Même lorsque l'estimation est possible, l'interprétation des résultats est donc extrêmement délicate. Il est alors indispensable de combiner à leur vraisemblance un critère supplémentaire qui régularise l'estimateur. Dans le sillage de la régression PLS, la stratégie de régularisation que nous considérons dans cette thèse est fondée sur l'extraction de composantes supervisées. Contraintes à l'orthogonalité entre elles, ces composantes doivent non seulement capturer l'information structurelle des variables explicatives, mais aussi prédire autant que possible les variables réponses, qui peuvent être de types divers (continues ou discrètes, quantitatives, ordinales ou nominales). La régression sur composantes supervisées a été développée pour les GLMs multivariés, mais n'a jusqu'alors concerné que des modèles à observations indépendantes.Or dans de nombreuses situations, les observations sont groupées. Nous proposons une extension de la méthode aux GLMMs multivariés, pour lesquels les corrélations intra-groupes sont modélisées au moyen d'effets aléatoires. À chaque étape de l'algorithme de Schall permettant l'estimation du GLMM, nous procédons à la régularisation du modèle par l'extraction de composantes maximisant un compromis entre qualité d'ajustement et pertinence structurelle. Comparé à la régularisation par pénalisation de type ridge ou LASSO, nous montrons sur données simulées que notre méthode non seulement permet de révéler les dimensions explicatives les plus importantes pour l'ensemble des réponses, mais fournit souvent une meilleure prédiction. La méthode est aussi évaluée sur données réelles.Nous développons enfin des méthodes de régularisation dans le contexte spécifique des données de panel (impliquant des mesures répétées sur différents individus aux mêmes dates). Deux effets aléatoires sont introduits : le premier modélise la dépendance des mesures relatives à un même individu, tandis que le second modélise un effet propre au temps (possédant donc une certaine inertie) partagé par tous les individus. Pour des réponses Gaussiennes, nous proposons d'abord un algorithme EM pour maximiser la vraisemblance du modèle pénalisée par la norme L2 des coefficients de régression. Puis nous proposons une alternative consistant à donner une prime aux directions les plus "fortes" de l'ensemble des prédicteurs. Une extension de ces approches est également proposée pour des données non-Gaussiennes, et des tests comparatifs sont effectués sur données Poissonniennes. / High redundancy of explanatory variables results in identification troubles and a severe lack of stability of regression model estimates. Even when estimation is possible, a consequence is the near-impossibility to interpret the results. It is then necessary to combine its likelihood with an extra-criterion regularising the estimates. In the wake of PLS regression, the regularising strategy considered in this thesis is based on extracting supervised components. Such orthogonal components must not only capture the structural information of the explanatory variables, but also predict as well as possible the response variables, which can be of various types (continuous or discrete, quantitative, ordinal or nominal). Regression on supervised components was developed for multivariate GLMs, but so far concerned models with independent observations.However, in many situations, the observations are grouped. We propose an extension of the method to multivariate GLMMs, in which within-group correlations are modelled with random effects. At each step of Schall's algorithm for GLMM estimation, we regularise the model by extracting components that maximise a trade-off between goodness-of-fit and structural relevance. Compared to penalty-based regularisation methods such as ridge or LASSO, we show on simulated data that our method not only reveals the important explanatory dimensions for all responses, but often gives a better prediction too. The method is also assessed on real data.We finally develop regularisation methods in the specific context of panel data (involving repeated measures on several individuals at the same time-points). Two random effects are introduced: the first one models the dependence of measures related to the same individual, while the second one models a time-specific effect (thus having a certain inertia) shared by all the individuals. For Gaussian responses, we first propose an EM algorithm to maximise the likelihood penalised by the L2-norm of the regression coefficients. Then, we propose an alternative which rather gives a bonus to the "strongest" directions in the explanatory subspace. An extension of these approaches is also proposed for non-Gaussian data, and comparative tests are carried out on Poisson data.

A semiparametric proportional likelihood ratio model was proposed by Luo and Tsai (2012) which is suitable for modeling a nonlinear monotonic relationship between the response variable and a covariate. Extending the generalized linear model, this model leaves the probability distribution unspecified but estimates it from the data. In this thesis, we propose to extend this model into analyzing the longitudinal data by incorporating random effects into the linear predictor. By using this model as the conditional density of the response variable given the random effects, we present a maximum likelihood approach for model estimation and inference. Two numerical estimation procedures were developed for response variables with finite support, one based on the Newton-Raphson algorithm and the other one based on generalized expectation maximization (GEM) algorithm. In both estimation procedures, Gauss-Hermite quadrature is employed to approximate the integrals. Upon convergence, the observed information matrix is estimated through the second-order numerical differentiation of the log likelihood function. Asymptotic properties of the maximum likelihood estimator are established under certain regularity conditions and simulation studies are conducted to assess its finite sample properties and compare the proposed model to the generalized linear mixed model. The proposed method is illustrated in an analysis of data from a multi-site observational study of prodromal Huntington's disease.