Finding a well-predicting model is one of the main goals of regression analysis. However, to evaluate a model's prediction abilities, it is a normal practice to use criteria which either do not serve this purpose, or criteria of insufficient reliability. As an alternative, there are relatively new methods which use repeated simulations for estimating an appropriate loss function -- prediction error. Cross-validation and bootstrap belong to this category. This thesis describes how to utilize these methods in order to select a regression model that best predicts new values of the response variable.

Introduction: The traditional IT investment evaluation methods and/or techniques tend to measure the quantitative value added by eHealth. However, there are contributions brought by innovation which are intangible and sundry, and thus are difficult to identify, measure and manage. A model presented by Vimarlund & Koch (2011) aims for identifying the benefits that IT investments bring to health and social care organizations. It could be used as a tool that identifies and classifies the effects and indicators of IT innovation in-vestments at different organizational levels for different stakeholders. Purpose and research questions: This is an evaluative study with the purpose to validate Vimarlund & Koch’s (2011) evaluation model through practical application. A care study of EIAS (ERAS Interactive Audit System) is conducted. ERAS stands for Enhanced Recovery After Surgery, which is an innovative process aims to enhance patient’s outcome after ma-jor surgery. EIAS is a system that supports the ERAS process. The aim is to achieve a deep understanding of IT investment evaluation. The model will be used in a real case as a guide to evaluate and identify impact that derives from the use and implementation of IT applica-tions. The process of evaluation could also be seen as a process of validation of the model in terms of comprehensiveness, practicality and applicability. Through this study, we aim to find out: 1) What are the possible contributions that EIAS brings to Jönköping County Council? 2) How is the performance of Vimarlund & Koch’s (2011) evaluation model in practical application, in terms of comprehensiveness, practicality and applicability? Method: The purpose of this study is evaluative and it is conducted by using adductive ap-proach. Single case study will be adopted as the research strategy. In this study, qualitative data will be collected through semi-structured interview with key respondents. The data collected will be analyzed qualitatively with a narrative approach. Conclusion: Guided by Vimarlund & Koch’s (2011) evaluation model, the innovations that have been brought into healthcare organizations by EARS are electronic information supply, internal integration of clinical information and possibilities to learn from the system. The model has been validated in terms of comprehensiveness, practicality and applicability. The evaluation model is a generic model to demonstrate the contribution of IT to innovation and change in health care. It could be used in both formative and summative assessment and as well as goal-free and goal-based evaluation. The issue of the productivity paradox has been noticed as some effects are not immediate after introducing of IT. User-participation or not could be considered as an important condition for the validity of the evaluation guided by the evaluation model.

Title: Uncertainties in regional climate models outputs Author: RNDr. Eva Holtanová Supervisor: doc. RNDr. Jaroslava Kalvová, CSc. Department: Dept. of Meteorology and Environment Protection Faculty of Mathematics and Physics Charles University in Prague Present doctoral thesis focuses on the analysis of uncertainties in regional climate model outputs in the area of the Czech Republic. Generally, the uncertainties in model outputs come from inaccuracies of initial and boundary conditions, further from the necessity to parameterize the small scale processes, and the structure of the model, e.g. the choice of numerical schemes or spatial resolution. In case of the simulations of future climate, another source of uncertainty arises. It is the unknown development of forcings that influence the climate system. The analysis in this work focuses on two multi-model ensembles, that come from two international European projects PRUDENCE and ENSEMBLES. The simulated 30-year mean seasonal air temperature and precipitation amounts are used, for the reference period 1961- 1990, and several future time periods. Two techniques were employed to assess the uncertainties. The first one was aimed at dividing the variance of a multi-model ensemble into contributions of regional model, driving global model and emission...

Most current automatic speaker recognition system extract speaker-depend features by looking at short-term spectral information. This approach ignores long-term information. I explored approach that use the fundamental frequency and energy trajectories for each speaker. This approach models prosody dynamics on single fonemes or syllables. It is known from literature that prosodic systems do not work as well the acoustic one but it improve the system when fusing. I verified this assumption by fusing my results with state of the art acoustic system from BUT. Data from standard evaluation campaigns organized by National Institute of Standarts and Technology are used for all experiments.

Agent-based simulation models are an important tool to study the effectiveness of policy interventions on the uptake of residential photovoltaic systems by households, a cornerstone of sustainable energy system transition. In order for these models to be trustworthy, they require rigorous validation. However, the canonical approach of validating emulation models through calibration with parameters that minimize the difference of model results and reference data fails when the model is subject to many stochastic influences. The residential photovoltaic diffusion model PVact features numerous stochastic influences that prevent straightforward optimization-driven calibration. From the analysis of the results of a case-study on the cities Dresden and Leipzig (Germany) based on three error metrics (mean average error, root mean square error and cumulative average error), this research identifies a parameter range where stochastic fluctuations exceed differences between results of different parameterization and a minimization-based calibration approach fails. Based on this observation, an approach is developed that aggregates model behavior across multiple simulation runs and parameter combinations to compare results between scenarios representing different future developments or policy interventions of interest.