(New!) Factor models for high dimensional mixed frequency time series (OeNB)
Project Leader: Em.O.Univ.Prof. Manfred Deistler
Assistants: Mag. Alexander Braumann, PhD Diego Eduardo Fresoli
Honest Confidence Sets for Sparsely and Non-Sparsely Tuned Model Selection Estimators (Deutsche Forschungsgemeinschaft)
Project leader: Assoc. Prof. Mag. Dr. Ulrike Schneider
Assistant: Bakk. Mag. Karl Ewald
In this project we want to investigate the distributional properties of shrinkage estimators, such as the popular Lasso estimator and other regularization methods, with the aim of deriving honest confidence sets based on the these estimators. These types of estimators have seen immense interest in the statistics literature in recent years. However, it is still largely unknown how to construct valid condence sets based on such estimators - a question that is both of theoretical as well as of practical interest.
Generalized Dynamic Factormodels - The Single and the Mixed Frequency Case
Project Leader: Em.O.Univ.Prof. Manfred Deistler
Assistant: Dipl.-Ing. Elisabeth Felsenstein, Dipl.-Ing. Lukas Kölbl , Mag. Alexander Braumann
The proposed research project deals with data-driven modelling of Generalized Linear Dynamic Factor Models (GDFM). Such models are used in particular to analyze and forecast high-dimensional time series. The importance of this area has increased considerably over the last decade. Our approach to this problem is based on system theory and its methods.
The project consists in the following parts:
- Single Frequency: We want to further extend our previous results obtained for the AR case. However, our emphasis will be on the more general ARMA case.
- Structure theory: The focus will be on the ARMA case, in particular on the development of a canonical form for ARMA systems which gives an AR representation whenever the process is AR. Moreover, it is intended to investigate the topological and geometric properties of parameter spaces and parameterizations.
- Estimation of (real-valued) AR and ARMA parameters: Emphasis is laid on the singular ARMA case for which naive (Gaussian) maximum-likelihood estimation is not possible. The Hannan-Rissanen procedure and appropriate modifications as well as the use of subspace procedures will be considered.
- Model Selection: Here we consider LASSO type estimation.
- Mixed Frequency: In many cases, in particular for high-dimensional time series, observations are available at different sampling frequencies. Our aim is to estimate the parameters of the system generating all data at the highest frequency from the mixed frequency data and to use this system for prediction, filtering and smoothing. Accordingly, a central issue will be to develop criteria for identifiability. If such a system is not identifiable from the mixed frequency data, we plan to develop alternative procedures for prediction, filtering and smoothing. Our idea is to work with so-called blocked systems. The estimation procedures developed for the single and mixed frequency case will be tested on real data, too.
Epileptic seizure propagation analysis
Project leader: Em.O.Univ. Prof. Manfred Deistler
Alumni: Dipl.-Ing. Dr. Christoph Flamm, Dipl.-Ing. Dr. Andreas Graef
A resective neurosurgical intervention can allow a successful treatment of therapy refractory patients. For the exact localization of the epileptic focus, EEG data are presurgically recorded and visually analyzed by neurologists. The analysis of these data is extremely difficult and time-intensive. The planned research project proposes an automated focus detection and seizure propagation analysis for epileptic seizures for the aforementioned presurgical monitoring by means of time series analysis methodology.
Based on invasive and scalp EEG data, spatial and temporal dependencies of brain areas shall be determined and analyzed during epileptic seizures. We want to aid physicians in the difficult interpretation of these EEG signals.
The planned research project comprises the following parts:
1. In a first step the temporal structure of the EEG recordings, which are highly instationary bio signals, shall be studied. Hereby, dynamic segmentation and methods for the detection of local stationarity are employed.
2. Based on these results, algorithms for reduction of the used input channels shall be developed. Methods like the An algorithm, which perform a dynamic channel selection, could be appropriate. Furthermore, by means of dynamic factor models we intend to model the fact that electrical signals originating from one source are recorded by several electrodes.
3. By means of graphical modeling and appropriate dependency measures coupling effects between the different electrodes shall be analyzed. The evolution of the obtained dependencies shall give a hint for the localization of the focus and the propagation of the seizure. Neurophysical informations shall be involved in the modeling.
4. We want to investigate the circumstances where scalp EEG data is sufficient for our analysis, and when invasive EEG data really yields relevant new information. Furthermore we want to compare the results of the EEG analysis with other investigation methods.
Modelling, forecasting, and control for econometrics based on generalized dynamic factor models: a system theoretic approach (Australian Research Council (ARC) Discovery Project)
Project leaders: Prof. Brian D.O. Anderson (ANU Canberra) and Em. Prof. Manfred Deistler
The project will provide a tool that will assist organizations wishing to understand the dynamics of a national economy to model it, and to forecast future econometric time series values. Such ability will provide another tool to econometric managers, including the Reserve Bank , Treasury and fund managers, that should benefit the Australian nation.
Banking Performance (OeNB) [completed]
Project leader: Ao.Univ. Prof. Wolfgang Scherrer
Assistant: Mag. Ronald Scheucher, BSc Oliver Leodolter
Over the last couple of years pressure on the banking system to reform and consolidate has increased significantly. Fueled initially by liberalization of financial markets the process has somewhat accelerated in the course of the still pending crisis of the global financial system. Still, despite of their effect on long-term stability of the financial system, operating inefficiency of banks and the resulting misallocation of resources throughout the financial system have still not got the necessary attention from regulatory bodies and policy makers.
To establish a case for a revision of sector policy we will produce and analyze efficiency levels for a balanced panel of European commercial banks over a ten-year period using advanced DEA (Data Envelopment Analysis) modeling and applying AHP (Analytic Hierarchy Process) to capture the strategic dimensions of financial and monetary policy in the European Union.
Building on DEA performance data we will try to model potential relations between efficiency in the banking sector and strategic policy variables and the macroeconomic trend, employing panel data models.
Our research into the efficiency of the European banking system is geared towards answering the following key questions:
1. How efficient are banks and the banking system in the Euro-zone?
2. How did efficiency develop over time (2002 to 2012)?
3. Why do overall efficiency levels perform the way they do?
4. How effective is current economic policy in the European Union in shaping operating performance in the banking sector?
Final report: OENB15495report.pdf
StratfüSys - Strategisches Führungssystem für die öffentlich-private Sicherheitszusammenarbeit [completed]
Project leader: Em.o.Univ. Prof. Manfred Deistler
Alumni: Mag. (FH) Gernd Watzenig, Mag.Phil. Leopold Schmertzing
Modeling and Forecasting Multivariate Time Series [completed]
Project leader: Em.O. Univ. Prof. Manfred Deistler
Alumni: Dipl.-Ing. Markus Waser, Dipl.-Ing. Christoph Flamm, Dipl.-Ing. Lukas Kölbl, Dipl.-Ing. Elisabeth Felsenstein
This proposal is concerned with modeling and forecasting of high dimensional time series, with a focus on dimension reduction. The following three areas are considered:
- Generalized linear dynamic factor models. Here the emphasis is laid on model specification (e.g. on estimation of the factor dimension), on selection of variables suited for factor analysis and on detection of zero patterns in the factor loading matrix (indicating factors which load only on part of the observations)
- Input selection and reduced rank linear dynamic models (in cases where the classification of the observations into inputs and outputs is known a priori)
- Graphical time series analysis, where causal relations between the variables are analysed
Industrial Diversity, Spatial Differentiation and Social Cohesion: Communicative Structures in the “Housing Topos” (OeNB) [completed]
Project Leader: Ao.Univ.Prof. Edwin Deutsch
Assistant: Martin Kerndler
This project is about the interplay between the spatial allocation of productive activities and housing services. The starting point is to test a number of hypotheses about spatial structures and their impacts on productivity and economic stability. We start from the basic hyporhesis that regions are more productive whenever they are characterized by a larger variety of economic activities. The study investigates if such structures can be identified from spatial economic data and if spatial spillover effects can be noted. With that we turn to the role of social cohesion in economic performance and ask, whether cohesion shows up in social interactions allowed for by the spatial composition of housing.
The object under study is the Austrian economy over the years 2003 to 2009. The data are drawn from the census, the Structural Business Statistics and other sources, which are aggregated to a time-spatial Panel on the meso-scale of NUTS3-units.
In a previous study, Deutsch and Wolf (2008) coined the notion of the “Housing Topos”, which denotes communicative structures in production and housing. The current project aims to expand the previous ideas in stringent spatial econometric models, which permit to test the hypotheses with reference to the literature.
Generalized Factor Models [completed]
Project Leader: Em. O. univ. Prof. Manfred Deistler
Alumni: Mag. Dr. Jasmin Sahbegovic, Dipl-Ing. Dr. Alexander Filler, Dipl.-Ing. Bernd Funovits, Dipl-Ing. Elisabeth Felsenstein
Generalized linear dynamic factor models are analysed. These models have been developed recently and they are used for analysis and forecasting of high dimensional time series in order to overcome the curse of dimensionality. We develop a structure theory, with emphasis on the zeroless case, which is generic in the setting considered. Accordingly the latent variables are modeled as a singular autoregressive process and (generalized) Yule Walker equations are used for estimation.
Structural Analysis of Labour Participation in the Life-Course [completed]
Project leader: Ao.Univ.Prof. Edwin Deutsch
Alumni: Andreas Wolf, Marcel Jira, Johanna Bertl
The project is a research study for the Austrian Ministry of Social Affairs. The objective is the statistical evaluation of the labour participation in the life-course, with emphasis on low income groups and on employment situations in risk-groups. Starting from the Austrian Census data over the horizon from 1989 to 2005 the research focuses on the derivation labour participation profiles in subsequent birth cohorts. By means of age-cohort models the project analyses the probabilities at which individuals of different ages are fully empoyed, part-time employed, marginally employed or unemployed. The method draws from age-cohort-period models in discrete choice. The analysis of the likely persistence of employment states in risk groups aims at supporting social policy decisions.
Commercial Estates in CEE-Countries: Forecast and Valuation [completed]
Project leader: Ao.Univ.Prof. Edwin Deutsch
Alumni: Andreas Wolf, Johanna Bertl
The project is financed by Bank-Austria Unicredit group. The objective is the forecast of economic input-data to control a scoring software, which evaluates real estate credits within the framework of Basel II. The project consists of an econometric analysis for the evaluation and forecast of commerce estate data in the CEE-countries Slovakia, Czech Republik and Hungary, with Austria as reference case, together with expertises for the development of a scoring-software. The scientific challenge is to set up time-series models for the relatively short-term data horizons of the CEE-data, and to determine the bandwith of the forecasts to control the scoring in a probabilistic sense. The project draws from the analytic background of Real Option theory approaches and from various stochastic models used to forecast financial data.