THE STUDENTS' SECTION OF OUR SITE!

Current PhD Students.

DI Christoph Aistleitner Investigation in Metric Discrepancy Theory. He is analyzing the discrepancy of sequences related to the sequence {n_k*alpha} by methods of probability theory. Especially important and unsolved is the case of sub-exponential and randomly generated n_k, e.g. when n_k is a random walk. The goal is to describe the relationship between the discrepancy of this sequence and its number theoretical properties. (Started: 10/06, supervisor: I. Berkes)

DI Johannes Schauer Statistical and Asymptotic properties of Augmented GARCH Processes. He is studying a new class of general augmented GARCH models with respect to the moments and stationarity conditions. The following questions are of interest: properties of strong approximation by Kiefer processes, a Wiener approximation of partial sums based on several functions and the parameter estimation of augmented GARCH processes by utilizing quasi maximum likelihood. (Started: 10/06, supervisor: I. Berkes)

Ahmad Basheer Samim Ansari Estimation Procedures for Generalized Linear Mixed Models. He is comparing the most common estimation procedures in generalized linear mixed models with respect to different structures of the random effects. The main focus lies on the observation of the standard error. A survey of existing software providing such estimation tools is given. Especially the routines of the package R will be treated. As a byproduct some new estimation routines for the specific models may be created. Besides a theoretical analysis he aimes to apply his results on proper data. The data set is based on several measurements at different locations and it provides the opportunity to investigate models with several very complex structured random effects. (Started: 9/05, supervisor: H. Friedl)

Dipl.-Math. Gordana Djuras Influence Factors for Word Length in Texts of Slavic Languages. She is studying texts of three slavic languages (Russian, Slovene, Croatian) with respect to possible influence factors for the word length which may allow for discrimination and clustering of texts. It should be analyzed which factors (author specific, text-typological and others) may be mainly responsible for word length. One aim is to assign a given text to a certain category (type of text, author) only with the knowledge of relevant statistical characteristics. This thesis is in co-operation with the Department of Slavic Studies, University of Graz (FWF-project P15485, 2002-2004: Prof. Peter Grzybek). (Started: 04/02, supervisor: E. Stadlober)

Mag. Nikolaus Haselgruber Reliability and Statistical Design of Experiments. For the development of technical components it is necessary to predict their reliability on the basis of measurements at the test bench, field trials and tests of the costumers. The load matrix method delivers estimations of reliability and predictions of lifetime with the help of accelerated testing and models of deterioration. The current approaches used in practise has to be checked for modifications and improvement. One goal is to investigate the usability and stability of statistical methods in case of applications where the number of observations in the experimental design is changing over time. This thesis is in co-operation with AVL-List GmbH (DI Dr. Klaus Denkmayr) (Started: 04/05, supervisor: E. Stadlober)

DI Dr. Johannes Hofrichter Time-dependent Models of Hydrological Data. He is studying change-point models for curves of ground water drains. Such a drain can be described as predictor of a generalized linear model (GLM). Change points can be established by different statistical approaches, e.g. hierarchical binary splitting and dynamic programming algorithms. The number of change points may be analyzed via some information criteria. Models for several drain curves with a given correlation structure will be investigated by random effects models.(Finished: 02/07, supervisor: H. Friedl)

Mag. Dr. Siegfried Hörmann Fluctuation analysis of dependent random processes. He extended the classical fluctuation theory of martingales and proved upper-lower class results under optimal conditions. He closed also the gap between independent theory and martingale theory. He obtained asymptotic results for ARCH type processes, a class of stationary processes important in modern econometrics. Additionally, he extended and sharpened a number of basic results in almost sure central limit theory, another recently investigated field of probability theory, exhibiting unusual dependance behavior. The proofs are based on strong approximation due to Strassen and applied to dependent processes by Philipp and Stout. In the new approach a direct a.s. approximation by independent random variables is applied under very general conditions which leads to optimal results. (Finished: 11/06, 147 p., supervisor: I. Berkes/ L. Horvath (University of Utah, Salt Lake City).

DI Dr. Sigrid Gerlinde Kern(Best thesis award in Applied Statistics of the Austrian Society for Statistics) Die stochastische Modellierung des EEX-Spotmarktes und die Bewertung von Swing-Optionen (Stochastic modeling of the EEX spot market and pricing of swing options). She developed a complex stochastic model for the hourly spot prices of the European Energy Exchange Market. Linear state space models with regime switching allow the modeling of price spikes considering crucial factors as the type of the day, period of the day, season of the year and long term trends. The quality of the approach is investigated by a thorough simulation study. Finally, the pricing model provides the basis for calculating the value of swing options. In particular, four different multi-exercise options (virtual power stations with reservoir) are analyzed. All algorithms used are implemented as Matlab programs. This thesis was in co-operation with the STEWEAG-STEG GmbH (Portfolio and Risk Management, MBA DI Dr. Gerald Supper) (Finished: 02/06, 157 p., supervisor: E. Stadlober/I. Berkes).

DI Dr. Bernd Genser  Analyse palliativer Therapiesequenzen beim metastasierten Mammakarzinom unter Anwendung von Multi-State Modellen (Analysis of palliative treatment sequences in metastatic breast cancer multi state models). Appendix. He analyzed data from a historical cohort consisting of 270 metastatic breast cancer patients. Using a multi-state approach the course of the patients during the different clinical progression states is modelled as a stochastic process. A set of common prognostic factors and the applied treatments are modelled as covariates. The partial Markovian structure of the observed stochastic process is the basic assumption to perform parameter estimation in the extended Cox-model. The software package S-Plus is used to demonstrate the implementation of marginal model approaches. The results of the multi-state models showed that the influence of prognostic factors in the metastatic phase of breast cancer is constant and independent of the applied treatment sequences. (Finished: 12/02, 134 p. supervisor: E. Stadlober/K.-D. Wernecke, Humboldt University Berlin)

DI Dr. Michael Kocher  High-dimensional data analysis for realistic electronic circuit simulation. He introduced three methods based on the statistical concept of location depth which is a special technique for finding boundary points and centrally located points in the high-dimensional parameter space. It opens the door to simulation models that allow both fast and realistic simulations. The procedures will be integrated in an automated generation flow to be applicable in a production and circuit design environment. The statistical methods are validated by simulation experiments of typical analog/mixed-signal circuit designs.This was a co-operative work with austriamicrosystems AG, an international company for the development and production of semiconductors (DI Dr. Gerhard Rappitsch). (Finished: 05/02, 167 p., supervisor: E. Stadlober/H. Friedl)

DI Dr. Erwin Stampfer  Prototypen von Verteilungen. He studied approximations of a random quantity X by a simpler one, Y = h(X), such that the mean squared error is minimal. The approximations are principal points, principal components and principal curves. For the problem of principal points several new methods of estimation are described and evaluated theoretically and by means of simulation. For curved data a suitable method of approximation is suggested which positions principal points on principal curves. The algorithms are implemented in SPlus and may be used for discretization, optimal grouping as well as clustering. (Finished: 06/00, 168p., supervisor: E. Stadlober/H. Friedl)

Mag. Dr. Franz Quehenberger  Schätzung von marginalen Modellen mit proportionalen Hasardfunktionen für Gruppen von korrelierten Zählprozessen. He discussed asymptotic properties of estimators of model parameters in the semiparametric Cox model adapted to clustered data. Two simulation studies showed essentially no differences between four alternate estimators. However,the estimator based on U-statistics were more sensitive against thinning of the risk set. The estimators are applied to data from dental medicine, whereby the durability of telescopic crowns clustered by patients is modelled. (Finished: 08/99, 195p., supervisor: E. Stadlober/W. Müller)

Birgit Sponer Extrahierung von Schlüsselindikatoren für die Projektsteuerung durch multivariate statistische Methoden (Extraction of Key Indicators for Project Control using multivariate statistical methods). She is working on the creation of statistical models, especially multiple regression models, for the expected values of project costs and project durations in all business areas of austriamicrosystems AG in Unterpremstätten. Therefore she develops definition, monitoring and empirical evaluation of the goodness of key indicators. Furthermore she works on the implementation of confidence intervals for average project costs and durations as well as she studies the analysis of temporal dependence of project parameters and the development of control regions for current projects. The results should be made available in form of a tool, programmed in R. This thesis is carried out in co-operation with austriamicrosystems AG, Unterpremstätten (DI Dr. Gerhard Rappitsch). (Start: 4/07, supervisor: E. Stadlober)

Esther Lichtenegger Statistische Versuchsplanung zur Optimierung der Parametrisierung von Klebeprozessen (Experimental design to optimize parameters of the splicing process). She is investigating fractional factorial designs which may help to improve the quality of splicing processes. The thesis contains the whole cycle of development: design and statistical analysis of the experiments, evaluation of the results with respect to applications in practise. This thesis is in co-operation with Magna Steyr Fahrzeugtechnik AG&Co KG, Graz (DI Wilhelm Trummer) (Started: 11/06, supervisor: E. Stadlober)

Christoph Augustin EDV-basierte Thermikkarte für Segelflieger (IT-based thermal map for glider pilots). He is analyzing past glider flights from given geographical regions to get information on weather conditions and locations on the ground that may have generated thermals. This should be investigated for different locations with respect to day time and season. A map should be constructed giving estimations of probabilities for thermals under certain light and wind conditions. This thesis is in co-operation with Naviter d.o.o., Celje, Slovenia (Started: 07/06, supervisor: E. Stadlober)
 

DI Katharina Stangl Versicherungsmathematische Analyse des Bonus-Malus Systems in der KfZ-Haftpflichtversicherung (Mathematical analysis of the bonus-malus-system in motor vehicle third party liability insurance) The thesis starts with a description of the austrian bonus-malus-system with its classification and its reclassification rules. In case of an accident an insured has two possibilities. Either he reports the damage to the insurance company and moves to a higher premium level, or he pays the damage by himself . Using the theory of stochastic dynamic programming threshold values are calculated in such a way, that the expected discounted costs of the insured are minimized. This threshold values are those damage levels from which on it is better for an insurant to claim. Especially for middle premium levels these threshold values turn out to be unexpectedly high. Furthermore the total expected finite and infinite horizon cost functions for the insurant as well as for the insurer are derived. These cost functions are important for the premium calculation of the insurance company.

DI Johannes Poglitsch Selective strategies for inspecting sewerage nets and prediction models for sanitation (Selektive Kanalinspektionsstrategien und Prognosemodelle für die Sanierung). He developed a proportional sampling design based on a stratification (15 strata) of the sewerage network in the city of Graz. Recent inspections of sewerages are analyzed and allocated to their stratifications to obtain an estimation of the condition of the sewerage network. Using this approach, a stochastic model based on Markov chains is developed. This model is used for forecasting the states of two specific materials. The scenario without sanitation illustrates the ageing process of the materials, whereas the model with sanitation provides a realistic prognosis for the condition of the materials within the next years. This thesis was in cooperation with the city of Graz, Department of Canalization (DI Werner Sprung) (Finished: 12/06, 87 p., supervisor: E. Stadlober)

DI Florian Kölbl Aggregation of AR(2) processes (Aggregation von AR(2) Prozessen). LS estimators of the classical AR(2) process are considered when the underlying variables are aggregated sums of independent random coefficient AR(2) models. The asymptotic of the corresponding statistics is established and we show that in general consistency is not fulfilled. A central limit theorem for the case that the number of aggregated terms is much larger than the number of observations is given. An estimation method for the parameters of random coefficients distributions is given and illustrated by some examples. (Finished: 04/06, 45 p., supervisor: I. Berkes)

DI Johannes Schauer(Best thesis award in Mathematical Statistics of the Austrian Society for Statistics) Statistische und asymptotische Eigenschaften von augmented GARCH Prozessen (The statistical and asymptotic properties of augmented GARCH processes). Large sample properties of the augmented GARCH(1,1) model were studied by Carrasco/Chen (2002) and Aue/Berkes/Horvath (2005), but little is known about augmented GARCH(p,q) sequences. A functional limit theorem for modified augmented GARCH processes is established which provides asymptotic distributions of functionals necessary for statistical inference in unit roots tests or change-point procedures. A simulation study illustrates the conditions of stationarity and the path behavior of GARCH models. (Finished: 04/06, 70 p., supervisor: I. Berkes)

DI Günther Sieghartsleitner Bewertung fondgebundener Versicherungsverträge mit garantiertem Mindestgewinn (Evaluation of insurance contracts bounded on funds where a minimum gain is guaranteed). Contracts for conservative investors include sometimes a guarantee for a minimum return of investment (e.g. via a guarantee for the yearly rate of interest g). If the fund performs better, then the investor should also get a certain amount a of the additional gain, where the residual gain is collected by the bank to cover the risk. The question is: which choices of a and g may lead to a fair contract? This problem has been studied under several scenarios for the development of the underlying fund. (Finished: 10/05, 129 p., supervisor: W. Müller)

DI Maximilian Schwaiger Vorhersage von Prozessparametern in der Styroporerzeugung (Prognosis of process parameters in Styropor production. The process of foaming expandable polystyrol (Styropor) is governed by different parameters. Based on data from production processes he found out tendencies and recipes which lead to optimal quality. The most important parameters proved to be the input weight, the steam pressure and the diameter of the raw material pearls. For a fast and efficient analysis of the corresponding data bank, a tool was programmed in MS Excel. This thesis was in cooperation with Hirsch Maschinenbah GmbH (DI Dr. Klaus Unterleutner). (Finished: 10/05, 84 p., supervisor: E. Stadlober)

DI Waltraud Richter Marginal versus conditional models for the analysis of time-dependent observations. She is dealing with generalized estimation equations (GEE) and random effects models (REMs). With GEES marginal expectations are specified whereas with REMs one is modelling conditional expectations. A comparison of these two approaches is carried out in combination with a practical data set based on a biometric study describing time-dependent measurements of brain masses. (Finished: 12/04, supervisor: H. Friedl)

DI Denise Pachernegg  Optimales Stoppen von Stochastischen Prozessen (Optimal stopping of stochastic processes). This thesis is founded on the deal with the modified forms of the classical Secretary Problem,which could be seen as a sequential problem of observation and choice, in which the payoff depends on the observations only through the relative range and not through the actual values. The theory of backward induction or dynamical programming can be considered, which leads in most cases to nonlinear differential equations. For some special problems the existence of an optimal stopping rule was proofed. Some modifications of the problem connected with game theory and some problems with known distribution are also discussed. (Finished: 01/04, 120 p., supervisor: I. Berkes)

DI Andreas Kvas  Multivariate statistische Prozesskontrolle in der Papierproduktion (Multivariate statistical process control in paper production). He established a software tool in C++ for multivariate process controlling. A special form of Hotelling's T² statistic is used for that purpose. The detailed construction of a historical data set (HDS) is discussed which has to be used as reference for observations from future manufacturing processes. The implementation of the software is described and tested by data from a paper production with 6 parameters. It is demonstrated that the procedure with the T² statistic as control variable leads to good results for the problem discussed. This thesis was in co-operation with Audit Software GmbH (DI Herbert Bischof).(Finished: 01/04, 92 p., supervisor: E. Stadlober)

DI Gregor Kvas  Statistische Versuchsplanung in der Papierproduktion (Experimental design in paper production.) Paper production is governed by key parameters which may influence the quality of the end product. Based on data from paper productions with different adjustments of the parameters it has to be studied (i) which parameters may responsible for roughness and printability and (ii) which specific adjustments may lead to good results in quality. The roughness of the surface of raw paper can be described by five parameters, and the roughness and printability of coated paper depend on four to six parameters. We were able to establish the bandwidth for parameter adjustments leading to best possible quality of raw paper as well as coated paper. This thesis was in co-operation with Audit Software GmbH (DI Herbert Bischof).(Finished: 01/04, 123 p.,supervisor: E. Stadlober)

DI Reinhard Fiedler  Die statistische Analyse und Prognose von Lastganglinien. (Statistical analysis and prognosis of load curves.) He compared samples of load curve profiles from styrian households with so called norm profiles and also investigated the spread of the input power. One goal was to establish statistical models for load curves of different consumer groups. For this SARIMA and SARIMAY models has been considered and compared with neural network models. This thesis was in co-operation with the STEWEAG-STEG GmbH (Portfolio and Risk Management, MBA DI Dr. Gerald Supper). (Finished: 03/03, 99 p., supervisor: E. Stadlober)

DI Renate Radaschitz  Die Anwendung der Scan Statistik auf Daten des Steirischen Fehlbildungsregisters (The application of scan statistics examplified by the Styrian Malformation Registry.) The Styrian register on dysplasia contains data from children up the the age of one year. Data from three categories are analyzed where the variation in population as well as different sizes of clusters were taken into account. Trends and relevant risk factors as the mother´s age has been identified. This thesis was in co-operation with the Department of Medical Informatics, Statistics and Documentation, University of Graz (Prof. Andrea Berghold). (Finished: 03/03, p. 76, supervisor: E. Stadlober)

DI Mario Djuzelic  Einflussfaktoren auf die Wortlänge und ihre Häufigkeitsverteilung am Beispiel von Texten slowenischer Sprache. The variable word length measured by the number of syllables, plays an important role in the statistical investigation of texts. He carried out a case study based on 153 slovene texts demonstrating the usefulness of discriminant analysis. These slovene texts are split into three different categories (literary prose, journalistic prose, poetry) and they are from different epochs and different authors. The procedure starts with a pool of variables as text length, mean, standard deviation and functions of higher moments of word length. Only a few (3 to 4) variables responsible for the discrimination of the texts remain for inclusion in the final discriminant function. The results indicate that similar questions tackled with this approach may lead to suitable answers. This thesis was in co-operation with the Department of Slavistic, University of Graz (Prof. Peter Grzybek). (Finished: 04/02, p. 100, supervisor: E. Stadlober)

DI Verena Meister  Testing of random number generators. She was discussing different approaches for random number generators (RNGs) and focused on the investigation of generators based on the linear congruential method. Their theoretical power, i.e. the structure of the underlying lattice was analyzed by the spectral test. Two generators with known lattice structure, Marsaglia´s Super-Duper generator with period length 2^32 and a generator with a larger period 2^52 were chosen for a thorough empirical analysis of the randomness of subsequences. The randomness was checked by a battery of statistical test implemented as a flexible system in C++. The results demonstrate that both proposals may be regarded as suitable RNGs. This thesis was in co-operation with ATRONIC, an international company producing electronic games for casinos. (Finished: 03/02, p. 126, supervisor: E. Stadlober/E. Stampfer)

DI Carmen Pleschiutschnig  Lattice points in bodies with algebraic boundary. She studied recent results on asymptotic values for the number of lattice points in ellipsoids. The first part of the thesis is concerned with a slightly improvement of a result due to Bentkus and Götze. The second part contains a recent result of W. Müller on the number of lattice points in bodies with algebraic boundaries. The corresponding proof is an application of the Hardy-Littlewood method. (Finished: 02/02, supervisor: W. Müller)

DI Egon Feiel  (Best thesis award in Applied Statistics of the Austrian Society for Statistics) Simulation und statistische Analyse von kreisförmigen und elliptischen Konturformen. He studied methods of shape statistics to estimate parameters of contour functions (circle, ellipse). The aim was to find out how many points of measurements are needed in estimating a contour function with a given precision on the characteristic parameters. Simulation studies and experiments with series of measurements delivered hints for procedures which may be integrated in the process of quality control. This thesis was in co-operation with TBK-Automatisierung, a small sized enterprise establishing units for electronic measurements. (Finished: 07/01, 99p., supervisor: E. Stadlober/E. Stampfer)

DI Eva Hajdu  Bestimmung relevanter Parameter zur Erkennung der Auslastung von VoIP Netzwerken. First, characteristics of efficiency and quality for Voice over Internet Protocol networks are defined and data over a period of 58 weeks are analysed. Second,the weekly medians of aggregated daily traffic values are fitted by ARIMA models. Third, the selected models are used for forecasts of a period of 18 weeks and these forecasts are compared with real data. This thesis was in cooperation with Infonova GmbH, a telecommunication company. (Finished: 07/01, 99p., supervisor: E. Stadlober)

DI Markus Eisner  Online Adjustment of D-optimal Designs. He investigated tools for optimal design generation and was responsible for evaluating their usability. The aim was to develop procedures for the on-line adaptation of test designs suitable in design spaces with certain restrictions. The resulting algorithms has been evaluated at testbeds and in a simulation environment. This thesis was in co-operation with AVL List GmbH, an international R&D company. (Finished: 06/01, 83p., supervisor: E. Stadlober/E. Stampfer)

DI Florian Kuss  Response Surface Models for the Calibration of Combustion Engines. He constructed algorithms based on standard tools of a MATLAB-library which choose sets of relevant parameters for describing response variables by polynomial regression equations. The resulting package contains the most recent tools of statistical modelling and will be integrated in a software package for the automatic calibration of powertrain management systems. This thesis is in co-operation with AVL List GmbH, an international R&D company. (Finished: 05/01, 94p., supervisor: E. Stadlober/E. Stampfer)

DI Verena Trattner  Die Analyse von Spurenelementen im Morgenharn steirischer Schüler. She studied the relations between concentrations on the corresponding bio-chemical parameters and their counterparts in soil and drink-water samples. The main goal of this study was to find out occasional differences in the heavy metal absorption of children at different measurement sites. The later were chosen carefully to represent the whole variety of regional territories with their different structures (like big cities, agricultural or mining industry regions) in Styria.(Finished: 05/01, 77p., supervisor: H. Friedl/E. Stampfer)

DI Andreas Zehetner  Statistische Analyse der Zugriffe auf die Homepage des Instituts für Statistik. He analysed the data of our Homepage from 01/98 to 12/99 to figure out typical patterns and characteristics of users as well as profiles of specific sites depending on different categories. He was able to identify gamma distributions as proper univariate models for the no. of accesses, and he looked at the no. of user sessions on a certain day t, leading to time series analysis. The obvious weekly fluctuations indicated the use of seasonal integrated autoregressive moving average models (SARIMA). A large number of possible SARIMA models have been assessed by using goodness-of-fit tests and information criteria. Two models of this type are selected and the corresponding values are used to forecast the data for the first quarter of 2000. It turns out that these special SARIMA-Models are able to predict the real data in an adequate manner. All the analysis has been carried out in S-Plus. (Finished: 09/00, 82p., supervisor: E. Stadlober/E. Stampfer)

DI Dina Ritter  Ordinal Models for the Analysis of Biotropy in Styria. She analysed a data set about biotropy levels, which were monitored during more than three years at eight different sites in Styria, and the potential influence of weather conditions. The main interest was about the necessity of all the eight stations to describe the behaviour of this bio-weather in our province. (Finished: 09/00, 92p., supervisor: H. Friedl/E. Stampfer)

DI Cornelia Brugger  Analyse der Zeitaufnahmeauswertung laut REFA nach statistischen Gesichtspunkten. She analysed two methods for confidence intervals of mean time periods and tested their robustness against departures from normality by simulations with gamma and lognormal data. These simulations and an example with real data showed that both methods yield reliable results for skewed distributions of sample sizes of at least 20.(Finished 07/00, 85 p., supervisor: E. Stadlober)

DI Harald Krainer  Statistische Auswertung von Charakterisierungsdaten integrierter Schaltungen. He explored test data obtained from automatic test procedures for integrated circuits. The aim was to analyse the so-called shmooplots and composite shmooplots to (i) get capability numbers, (ii) to detect systematic defects and to (ii) evaluate the influence of temperature on the results during the characterization of the ICs. The resulting algorithms were embedded into an existing system of programs. The work was done in in co-operation with Infineon technologies, Graz. (Finished: 06/00, 102 p., supervisor: E. Stadlober/E. Stampfer)

DI Johannes Hofrichter  Residuen-Resampling mit S-Plus. He compared various residual resampling strategies within the class of Generalised Linear Models. A macro-package was developed which allows for calculating the bootstrap variances as well as their Monte-Carlo approximations. The special interest was on a comparison between resampling from Pearson residuals and resampling from deviance components. (Finished 9/1999, 70 p., supervisor: H. Friedl/E. Stampfer)

DI Albert Tösch  Korrespondenzanalyse. He investigated the usefulness of the so called Correspondence Analysis (CA). CA is an explorative method for analyzing (possibly large) contingency tables. Its main theme was to visuale the structure of the data. (Finished: 06/99, supervisor: H. Friedl/E. Stampfer)

DI Maximilian Gartner  Analyse von Lastganglinien mittels Extremwert- und Verallgemeinerter Pareto-Modelle. He analysed load curves of daytime dependent electricity rates using data on the power consumption of 400 customers. The following problems has been tackled: (I) determination of typical consumption curves for summer/winter and three different types of consumers, (II) statistical analysis of extreme values and exceedances. (Finished 03/99, 94 p. supervisor: E. Stadlober/E. Stampfer)

DI Erwin Fugger  Ein Qualitätskontrollsystem bei der Produktion von Halbleitern. He developed a software prototype for quality control in integrated circuit fabrication. His work was done for the manufacturer Siemens at Villach. (Finished 03/98, 104 p., supervisor: E. Stadlober)

DI Herbert Bischof  Modellierung und Vergleich spirometrischer Parameter von Rauchern und Nichtrauchern. He analysed the influence of smoking habits to lung function parameters. His investigations are based on the data of more than 10.000 female and male subjects. (Finished 10/97, 162 p., supervisor: E. Stadlober/H. Friedl)

DI Stefan Gesslbauer  Die Simulation von Verteilungen durch MCMC-Methoden mit adaptiver Verwerfung. He studied Markov-Chain-Monte-Carlo methods and was looking for the efficiency of adaptive rejection schemes for sampling from general continuous distributions. (Finished 10/97, 101 p., supervisor: E. Stadlober/H. Friedl)

DI Gunther Glawar  Statistische Evaluierung des Emissionskatasters der Stadt Graz. He estimated the amount of yearly emissions in the area of Graz and compared it with the results of other methods. This was based on data from companies of different branches obtained by a questionnaire. (Finished 10/97, 107 p., supervisor: E. Stadlober/H. Friedl)

DI Günter Blümel  Die Lognormalverteilung als Modell zur Schätzung von betrieblichen Energieverbrauchsfaktoren. He used the lognormal distribution to estimate energy consumption and emission. Data on energy consumption and emission of 311 out of a total of 1811 enterprises in Graz were used to estimate the consumption of the remaining 1500 enterprises. (Finished 08/97, 104 p., supervisor: E. Stadlober/H. Friedl)