Water balance models are important for the characterization of hydrologic systems, to help understand regional scale dynamics, and to identify hydro-climatic trends and systematic biases in data. Because existing models have, to-date, only been tested on data sets of limited spatial representativeness and extent, it has not yet been established that they are capable of reproducing the range of dynamics observed in nature. This dissertation develops systematic strategies to guide selection of water balance models, establish data requirements, estimate parameters, and evaluate performance. Through a series of three papers, these challenges are investigated in the context of monthly water balance modeling across the conterminous United States. The first paper reports on an initial diagnostic iteration to evaluate relevant components of model error, and to examine details of its spatial variability. We find that to conduct a robust model evaluation it is not sufficient to rely upon conventional NSE and/or r^2aggregate statistics of performance; to have reasonable confidence that the model can provide hydrologically consistent simulations, it is also necessary to examine measures of water balance and hydrologic variability. The second paper builds upon the results of the first, and evaluates the suitability of several candidate model structures, focusing specifically snow-free catchments. A diagnostic Maximum-Likelihood model evaluation procedure is developed to incorporate the notion of `Hydrological Consistency' and controls for structural complexity. The results confirm that the evaluation of hydrologic consistency, based on benchmark comparisons and on stringent analysis of residuals, provides a robust basis for guiding model selection. The results reveal strong spatial persistence of certain model structures that needs to be understood in future studies. The third paper focuses on understanding and improving the procedure for constraining model parameters to provide hydrologically consistent results. In particular, it develops a penalty-function based modification of the Mean Squared Error estimation to help ensure proper reproduction of system behaviors by minimizing interaction of error components and by facilitating inclusion of relevant information. The analysis and results provide insight into the identifiability of model parameters, and further our understanding of how performance criteria should be applied during model identification.

Complete protection against flood risks by structural measures is impossible. Therefore flood prediction is important for flood risk management. Good explanatory power of flood models requires a meaningful representation of bio-physical processes. Therefore great interest exists to improve the process representation. Progress in hydrological process understanding is achieved through a learning cycle including critical assessment of an existing model for a given catchment as a first step. The assessment will highlight deficiencies of the model, from which useful additional data requirements are derived, giving a guideline for new measurements. These new measurements may in turn lead to improved process concepts. The improved process concepts are finally summarized in an updated hydrological model. In this thesis I demonstrate such a learning cycle, focusing on the advancement of model evaluation methods and more cost effective measurements. For a successful model evaluation, I propose that three questions should be answered: 1) when is a model reproducing observations in a satisfactory way? 2) If model results deviate, of what nature is the difference? And 3) what are most likely the relevant model components affecting these differences? To answer the first two questions, I developed a new method to assess the temporal dynamics of model performance (or TIGER - TIme series of Grouped Errors). This method is powerful in highlighting recurrent patterns of insufficient model behaviour for long simulation periods. I answered the third question with the analysis of the temporal dynamics of parameter sensitivity (TEDPAS). For calculating TEDPAS, an efficient method for sensitivity analysis is necessary. I used such an efficient method called Fourier Amplitude Sensitivity Test, which has a smart sampling scheme. Combining the two methods TIGER and TEDPAS provided a powerful tool for model assessment. With WaSiM-ETH applied to the Weisseritz catchment as a case study, I found insufficient process descriptions for the snow dynamics and for the recession during dry periods in late summer and fall. Focusing on snow dynamics, reasons for poor model performance can either be a poor representation of snow processes in the model, or poor data on snow cover, or both. To obtain an improved data set on snow cover, time series of snow height and temperatures were collected with a cost efficient method based on temperature measurements on multiple levels at each location. An algorithm was developed to simultaneously estimate snow height and cold content from these measurements. Both, snow height and cold content are relevant quantities for spring flood forecasting. Spatial variability was observed at the local and the catchment scale with an adjusted sampling design. At the local scale, samples were collected on two perpendicular transects of 60 m length and analysed with geostatistical methods. The range determined from fitted theoretical variograms was within the range of the sampling design for 80% of the plots. No patterns were found, that would explain the random variability and spatial correlation at the local scale. At the watershed scale, locations of the extensive field campaign were selected according to a stratified sample design to capture the combined effects of elevation, aspect and land use. The snow height is mainly affected by the plot elevation. The expected influence of aspect and land use was not observed. To better understand the deficiencies of the snow module in WaSiM-ETH, the same approach, a simple degree day model was checked for its capability to reproduce the data. The degree day model was capable to explain the temporal variability for plots with a continuous snow pack over the entire snow season, if parameters were estimated for single plots. However, processes described in the simple model are not sufficient to represent multiple accumulation-melt-cycles, as observed for the lower catchment. Thus, the combined spatio-temporal variability at the watershed scale is not captured by the model. Further tests on improved concepts for the representation of snow dynamics at the Weißeritz are required. From the data I suggest to include at least rain on snow and redistribution by wind as additional processes to better describe spatio-temporal variability. Alternatively an energy balance snow model could be tested. Overall, the proposed learning cycle is a useful framework for targeted model improvement. The advanced model diagnostics is valuable to identify model deficiencies and to guide field measurements. The additional data collected throughout this work helps to get a deepened understanding of the processes in the Weisseritz catchment. / Modelle zur Hochwasservorhersage und –warnung basieren auf einer bio-physikalisch Repräsentation der relevanten hydrologischen Prozesse. Eine Verbesserungen der Beschreibung dieser Prozesse kann zuverlässigere Vorhersagen ermöglichen. Dazu wird die Benutzung eines Lernzykluses bestehend aus einer kritische Beurteilung eines existierenden Modells, der Erhebung zusätzlicher Daten, der Bildung eines vertieften Verständnis und einer Überarbeitung des Modells vorgeschlagen. In dieser Arbeit wird ein solcher Lernzyklus aufgegriffen, wobei der Schwerpunkt auf einer verbesserten Modellanalyse und kosteneffizientere Messungen liegt. Für eine erfolgreiche Modellbeurteilung sind drei Fragen zu beantworten: 1) Wann reproduziert ein Modell die beobachteten Werte in einer zufriedenstellenden Weise (nicht)? 2) Wie lassen sich die Abweichungen charakterisieren? und 3) welches sind die Modellkomponenten, die diese Abweichungen bedingen? Um die ersten beiden Fragen zu beantworten, wird eine neue Methode zur Beurteilung des zeitlichen Verlaufs der Modellgüte vorgestellt. Eine wichtige Stärke ist, dass wiederholende Muster ungenügender Modellgüte auch für lange Simulationsläufe einfach identifiziert werden können. Die dritte Frage wird durch die Analyse des zeitlichen Verlaufs der Parametersensitivität beantwortet. Eine Kombination der beiden Methoden zur Beantwortung aller drei Fragen stellt ein umfangreiches Werkzeug für die Analyse hydrologischer Modelle zur Verfügung. Als Fallstudie wurde WaSiM-ETH verwendet, um das Einzugsgebiet der wilden Weißeritz zu modellieren. Die Modellanalyse von WaSiM-ETH hat ergeben, dass die Schneedynamik und die Rezession während trockener Perioden im Spätsommer und Herbst, für eine Beschreibung der Prozesse an der Weißeritz nicht geeignet sind. Die Erhebung zusätzlicher Daten zum besseren Verständnis der Schneedynamik bildet den nächste Schritt im Lernzyklus. Daten über Schneetemperaturen und Schneehöhen wurden mit Hilfe eines neuen, preisgünstigen Verfahrens erhoben. Dazu wurde die Temperatur an jedem Standort mit unterschiedlichen Abständen zum Boden gemessen und mit einem neuen Algorithmus in Schneehöhe und Kältegehalt umgerechnet. Die Schneehöhe und Kältegehalt sind wichtige Größen für die Vorhersage von Frühjahrshochwassern. Die räumliche Variabilität der Schneedecke auf der Einzugsgebietsskala wurde entsprechend der Landnutzung, der Höhenzone und der Ausrichtung stratifiziert untersucht, wobei lediglich der Einfluss der Höhe nachgewiesen werden konnte, während Ausrichtung und Landnutzung keinen statistisch signifikanten Einfluss hatten. Um die Defizite des WaSiM-ETH Schneemodules für die Beschreibung der Prozesse im Weißeritzeinzugsgebiets besser zu verstehen, wurde der gleiche konzeptionelle Ansatz als eigenständiges, kleines Modell benutzt, um die Dynamik in den Schneedaten zu reproduzieren. Während dieses Grad-Tag-Modell in der Lage war, den zeitlichen Verlauf für Flächen mit einer kontinuierlichen Schneedecke zu reproduzieren, konnte die Dynamik für Flächen mit mehreren Akkumulations- und Schmelzzyklen im unteren Einzugsgebiet vom Modell nicht abgebildet werden. Vorschläge zur Verbesserung des Modells werden in der Arbeit gemacht. Zusammenfassend hat sich das Lernzyklus-Konzept als nützlich erwiesen, um gezielt an einer Modellverbesserung zu arbeiten. Die differenzierte Modelldiagnose ist wertvoll, um Defizite im Modellkonzept zu identifizieren. Die während dieser Studie erhobenen Daten sind geeignet, um ein verbessertes Verständnis der Schnee-Prozesse an der Weißeritz zu erlangen.

Discrete choice models form a class of models widely used in econometrics for modelling the individual choice from a finite set of alternatives. The most widely used model is the multinomial logit model, implicitly assuming independence of irrelevant alternatives. A generalization is the nested multinomial logit model, relaxing this strong assurnp tion. Viewing both models as nonlinear regression models a set of diagnostics is derived. This includes a hat matrix, measures of leverage, influence and residuals and an approximation to the parameters for case deletion. In an example for the multinomid logit model a good performance of these diagnostics is observed and the parameter approximation by the proposed formula is better than a one step Newton-Raphson procedure. In an example for the nested logit model a constructed outlier with high influence is revealed by the measures of leverage and residual, but the parameter approximation is insufficient. (author's abstract) / Series: Forschungsberichte / Institut für Statistik

<p>Pharmacometric modelling is an increasingly used method for analysing the outcome from clinical trials in drug development. The model building process is complex and involves testing, evaluating and diagnosing a range of plausible models aiming to make an adequate inference from the observed data and predictions for future studies and therapy. </p><p>The aim of this thesis was to advance the approaches used in pharmacometrics by introducing improved models and methods for application in essential parts of model building procedure: (i) structural model development, (ii) stochastic model development and (iii) model diagnostics. </p><p>As a contribution to the structural model development, a novel flexible structural model for drug absorption, a transit compartment model, was introduced and evaluated. This model is capable of describing various drug absorption profiles and yet simple enough to be estimable from data available from a typical trial. As a contribution to the stochastic model development, three novel methods for parameter distribution estimation were developed and evaluated; a default NONMEM nonparametric method, an extended grid method and a semiparametric method with estimated shape parameters. All these methods are useful in circumstances when standard assumptions of parameter distributions in the population do not hold. The new methods provide less biased parameter estimates, better description of variability and better simulation properties of the model. As a contribution to model diagnostics, the most commonly used diagnostics were evaluated for their usefulness. In particular, diagnostics based on individual parameter estimates were systematically investigated and circumstances which are likely to misguide modelers towards making erroneous decisions in model development, relating to choice of structural, covariate and stochastic model components were identified. </p><p>In conclusion, novel approaches, insights and models have been provided to the pharmacometrics community. </p><p>Implementation of these advances to make model building more efficient and robust has been facilitated by development of diagnostic tools and automated routines.</p>

Pharmacometric modelling is an increasingly used method for analysing the outcome from clinical trials in drug development. The model building process is complex and involves testing, evaluating and diagnosing a range of plausible models aiming to make an adequate inference from the observed data and predictions for future studies and therapy. The aim of this thesis was to advance the approaches used in pharmacometrics by introducing improved models and methods for application in essential parts of model building procedure: (i) structural model development, (ii) stochastic model development and (iii) model diagnostics. As a contribution to the structural model development, a novel flexible structural model for drug absorption, a transit compartment model, was introduced and evaluated. This model is capable of describing various drug absorption profiles and yet simple enough to be estimable from data available from a typical trial. As a contribution to the stochastic model development, three novel methods for parameter distribution estimation were developed and evaluated; a default NONMEM nonparametric method, an extended grid method and a semiparametric method with estimated shape parameters. All these methods are useful in circumstances when standard assumptions of parameter distributions in the population do not hold. The new methods provide less biased parameter estimates, better description of variability and better simulation properties of the model. As a contribution to model diagnostics, the most commonly used diagnostics were evaluated for their usefulness. In particular, diagnostics based on individual parameter estimates were systematically investigated and circumstances which are likely to misguide modelers towards making erroneous decisions in model development, relating to choice of structural, covariate and stochastic model components were identified. In conclusion, novel approaches, insights and models have been provided to the pharmacometrics community. Implementation of these advances to make model building more efficient and robust has been facilitated by development of diagnostic tools and automated routines.

Model diagnostics for normal and non-normal state space models is based on recursive residuals which are defined from the one-step ahead predictive distribution. Routine calculation of these residuals is discussed in detail. Various tools of diagnostics are suggested to check e.g. for wrong observation distributions and for autocorrelation. The paper also covers such topics as model diagnostics for discrete time series, model diagnostics for generalized linear models, and model discrimination via Bayes factors. (author's abstract) / Series: Forschungsberichte / Institut für Statistik

Several sophisticated techniques to study in vivo GI transit and regional absorption of pharmaceuticals are available and increasingly used. Examples of such methods are Magnetic Marker Monitoring (MMM) and local drug administration with remotely operated capsules. Another approach is the paracetamol and sulfapyridine double marker method which utilizes observed plasma concentrations of the two substances as markers for GI transit. Common for all of these methods is that they generate multiple types of observations e.g. tablet GI position, drug release and plasma concentrations of one or more substances. This thesis is based on the hypothesis that application of mechanistic nonlinear mixed-effect models could facilitate a better understanding of the interrelationship between such variables and result improved predictions of the processes involved in oral absorption. Mechanistic modeling approaches have been developed for application to data from MMM studies, paracetamol and sulfapyridine double marker studies and for linking in vitro and in vivo drug release. Models for integrating information about tablet GI transit, in vivo drug release and drug plasma concentrations measured in MMM studies was outlined and utilized to describe drug release and absorption properties along the GI tract for felodipine and the investigational drug AZD0837. A mechanistic link between in vitro and in vivo drug release was established by estimation of the mechanical stress in different regions of the GI tract in a unit equivalent to rotation speed in the in vitro experimental setup. The effect of atropine and erythromycin on gastric emptying and small intestinal transit was characterized with a semi-mechanistic model applied to double marker studies in fed and fasting dogs. The work with modeling of in vivo drug absorption has highlighted the need for, and led to, further development of mixed-effect modeling methodology with respect to model diagnostics and the handling of censored observations.

1 AbstraktEN The diploma thesis deals with linear mixed effects models. In the first chap- ter, we discuss parameter estimation and hypothesis testing in the linear mixed effects models. The second chapter is dedicated to graphical diagnostics. We look at the suitable diagnostic plots for residuals and random effects estimates. It is closely described, how the violations of assumptions affect the diagnostic plots. In the third chapter we have consequences of the violations of assumptions on the parameter estimates and results of hypothesis testing for fixed effects. 1

<p>Doktorska disertacija ukazuje na model dijagnostike stanja koji zbog svog istraživačkog karaktera može dovesti do novih naučnih saznanja i metoda praćenja uticaja najvažnijih parametara na pouzdanost vozila, izučavanja ključnih performansi iz oblasti održavanja motornih vozila. Njen cilj je da teorijski i empirijski, kritički, sistematski i kontrolisano defini&scaron;e model dijagnostike stanja kao i da izvr&scaron;i izbor optimalnih parametara, radne temperature i pohabanosti ležajeva, a sve u cilju određivanja sigurnosti funkcionisanja sastavnih komponenti motornih vozila. Proces istraživanja modela predstavlja vezu između periodičnosti provere parametara stanja u radu i otkaza sastavnih komponenata motornih vozila. Simulacijom se može prognozirati vremenski trenutak zamene komponenata pre nego &scaron;to dođe do njihovog otkaza. Dati model je univerzalnog tipa iz razloga &scaron;to se može primeniti i na složene sisteme, bez obzira na dimenzije komponenti sklopova motornih vozila.</p> / <p>PhD dissertation indicates a model of state diagnostics, which due to its research nature, could lead to new scientific knowledge and methods of monitoring the impact of the most important parameters on vehicle reliability, the study of key performance in the field of maintenance of motor vehicles. Its aim is to theoretically and empirically, critically, systematically and in a controlled way define the model of conditions diagnostic and to make the selection of optimal parameters, operating temperature and wear of bearings, all for the purpose of determining the security of functioning of the parts and components of motor vehicles. The research process of a model represents the relationship between the periodicity of testing parameters of the operating mode and cancellation of integral components of motor vehicles. The simulation can predict time for replacement of components before they cancel. The present model is of a universal type because it can be applied to complex systems, regardless of the dimensions of the components of motor vehicles.</p>