Gastvortrag von Prof. Dr. Philipp Sibbertsen (Leibniz Universität Hannover) am 19. November 2018
Mo., 19. November 2018
13:00 Uhr, HG 217
Am 19. November wird Prof. Dr. Philipp Sibbertsen von der Leibniz Universität Hannover einen Gastvortrag zum Thema "Multivariate Spurious Long Memory and a robust Local Whittle Estimator" halten.
For univariate time series it is well documented that low frequency contaminations generate spurious long memory although both features are empirically distinct. This paper extends this analysis to vector valued processes. We propose a rigorous definition of spurious fractional cointegration, we show that such a behavior will occur in processes with joint low frequency contaminations and standard estimation of the cointegrating rank can spuriously indicate fractional cointegration in these situations.
To deal with multivariate low frequency contamination, we derive a robust multivariate local Whittle (RMLW) estimator for the memory parameters and the cointegrating vector that is consistent and asymptotically normal in the presence of low frequency contaminations and spurious fractional cointegration.
Finally, we introduce a procedure to consistently estimate whether there is fractional cointegration and common low frequency contaminations in the series.
Gastvortrag von Prof. Dr. Christian Weiß (Helmut-Schmidt-Universität Hamburg) am 29. November 2018
Do., 29. November 2018
13:00 Uhr, HG 217
Am 29. November wird Prof. Dr. Christian Weiß von der Helmut-Schmidt-Universität Hamburg einen Gastvortrag zum Thema "SPC methods for time-dependent processes of counts" halten.
In many fields of application, we are concerned with count processes. Typical examples are counts of defects per produced item in manufacturing industry, counts of new cases of an infection per time unit in health care monitoring, or counts of complaints by customers per time unit in service industry. Often, it is important to detect changes in the process as soon as possible to be able to start preventive actions or to avoid further damages. Methods of statistical process control are a suitable tool for this purpose.
During the last few years, there was increasing interest in SPC methods for time-dependent processes of counts. The talk surveys recent developments in this field. Some feasible models for autocorrelated count processes are briefly discussed, and approaches for corresponding control charts are considered. These cover the basic Shewhart chart as well as advanced control charts like CUSUM and EWMA methods. Also the topic of performance evaluation is briefly considered.
Gastvortrag von Prof. Dr. Timo Schmid (Freie Universität Berlin) am 28. Juni 2018
Do., 28. Juni 2018
14:00 - 15:00 Uhr, HG 217
Am 28. Juni hat Prof. Dr. Timo Schmid von der Freien Universität Berlin einen Gastvortrag zum Thema "Constructing socio-demographic indicators using mobile phone data: estimating literacy rates in Senegal" gehalten.
Modern systems of official statistics require the accurate and timely estimation of socio-demo-graphic indicators for disaggregated geographical regions. Traditional data collection methods such as censuses or household surveys impose great financial and organizational burdens for National Statistical Institutes. The rise of new information and communication technologies offers promising sources to mitigate these shortcomings.
In this paper we propose a unified approach for National Statistical Institutes based on small area estimation that allows for the estimation of socio-demographic indicators by using mobile phone data. In particular, the methodology is applied to mobile phone data from Senegal for deriving sub-national estimates of the share of illiterates disaggregated by gender. The estimates are used to identify hot spots of illiterates with a need for additional infrastructure or policy adjustments. Although the paper focuses on literacy as a particular socio-demographic indicator, the proposed approach is applicable to indicators in general.
For further information we refer to "https://rss.onlinelibrary.wiley.com/doi/epdf/10.1111/rssa.12305"
Gastvortrag von Prof. Dr. Roland Fried (Technische Universität Dortmund) am 07. Juni 2018
Do., 07. Juni 2018
14:00 - 15:00 Uhr, HG 217
Am 07. Juni hat Prof. Dr. Roland Fried von der Technischen Universität Dortmund einen Gastvortrag zum Thema "Robust time series analysis – the R-package robts" gehalten.
Many traditional methods of time series analysis are based on conditional least squares, Gaussian likelihood or empirical sample autocorrelations. Estimators and tests arising from these concepts can easily be misled by outliers and are not optimal in case of heavy-tailed distributions. Robust methods have been suggested in the literature therefore, particularly within the framework of linear time series modeling using, e.g., autoregressive moving average models. However, to the best of our knowledge, open-source software on such methods is scarce. Our R-package robts aims at filling this gap and provides functions which resemble the standard functions commonly used for analyzing autocorrelation or partial autocorrelation functions, for fitting autoregressive moving average time series models, for prediction and spectral density estimation. To achieve robustness we use M-estimators with bounded psi-functions instead of least squares or Gaussian likelihoods, and robust covariance and correlation estimators instead of ordinary sample autocorrelations. Change-point tests based on robust estimators allow us to check the basic assumption of 2nd order stationarity. The presentation illustrates the methods and our implementations using real data examples and studies their behavior via simulations. Our R-package comprises methods and techniques proposed by several authors including Maronna et al. . A review on robust estimators of autocorrelations and partial autocorrelations underlying our R-package can be found in Dürre et al. .
- Dürre, A., Fried, R. & Liboschik, T. (2015). Robust estimation of (partial) autocorrelation. WIREs Comput. Stat. 7, 205–222.
- Maronna, R., Martin, D. & Yohai, V. (2006). Robust Statistics: Theory and Methods. John Wiley & Sons, Chichester.
Gastvortrag Prof. Dr. Thomas Kneib (Universität Göttingen) am 30. November 2017
Do., 30. November 2017
14:15 - 17:45 Uhr (Raum HG 162)
Am 30.11.2017 hat Prof. Thomas Kneib einen Vortrag zu seiner aktuellen Forschung gehalten.
Semiparametric regression models offer considerable flexibility concerning the specification of additive regression predictors including effects as diverse as nonlinear effects of continuous covariates, spatial effects, random effects, or varying coefficients. Recently, such flexible model predictors have been combined with the possibility to go beyond pure mean-based analyses by specifying regression predictors on potentially all parameters of the response distribution in a distributional regression framework. In this talk, we introduce a generic concept for defining interaction effects in such semiparametric distributional regression models based on tensor products of main effects and illustrate this along the special case of spatio-temporal interactions. These interactions can be anisotropic, i.e. different amounts of smoothness will be associated with spatial and temporal effects. We investigate identifiability and the decomposition of interactions into main effects and pure interaction effects (similar as in a smoothing spline analysis of variance) to facilitate a modular model building process. Inference is based on Markov chain Monte Carlo simulations with iteratively weighted least squares proposals under constraints to ensure identifiability and effect decomposition.
Gastvortrag Prof. Dr. Alessandro Fassò (Universität Bergamo) am 26. Oktober 2017
Fr., 27. Oktober 2017
10:00 - 12:00 Uhr (GD 312)
Am 27. Oktober hat Prof. Alessandro Fassò von der Universität Bergamo einen Gastvortrag zum Thema "Statistical harmonization and uncertainty assessment in the comparison of satellite and radiosonde climate variables" gehalten.
Satellite validation is key to ensure the delivery of consistent long-term climate data records for climate and weather studies. Since the agreement of satellite measurements (SAT) with ground-based reference measurements (GND) is an essential quality indicator, one of the major issues to perform a rigorous validation is the quantification of the uncertainty due to the co-location mismatch in time and space between satellite and ground based reference observations.
This mismatch is due to the different sampling of atmosphere carried out by the two instruments (Verholest et al., 2015), which are also quite often based on very different remote sensing techniques. Moreover, satellite and ground based observations are typically collected on very different time and spatial scales. As a consequence, in a SAT-GND comparison we may have horizontal and/or vertical and/or temporal mismatches in smoothing and/or resolution, which are factors of the co-location mismatch uncertainty (hereinafter mismatch uncertainty). In the frame of Horizon 2020 GAIA-CLIM project (www.gaia-clim.eu), this paper considers the mismatch uncertainty in the comparison of the satellite observations obtained by the Infrared Atmospheric Sounding Interferometer (IASI) instrument with the radiosonde observations of the RAOB network.
Along these lines, the present paper focuses on the vertical smoothing mismatch uncertainty of IASI-RAOB profile comparison using a rigorous likelihood approach which takes into account autocorrelation, heteroscedasticity and atmospheric smoothness. In particular the vertical sparseness of RAOB network is assessed by means of a comparison with GRUAN reference products where available.
The proposed technique is a two-step technique. At the first step RAOB profiles are transformed into continuous functions using smoothing weighted splines, which are optimized to much as close as possible to GRUAN reference profiles. In doing this, vertical sparseness uncertainty and processing mismatch uncertainty are assessed. At the second step RAOB are harmonized by considering weighting functions based on the Generalized Extreme Values (GEV) probability density function (pdf) whose parameters are given by a regression on some covariates, which is regularized to take into account atmospheric smoothness. Since the model is highly nonlinear and data are not Gaussian, confidence intervals of parameter estimates and of mismatch uncertainties are assessed by a bootstrap technique.