|
|
|
| 2000 | Diploma (Engineering and Business Administration), Politehnica University, Bucharest, Romania | | 2001 - 2006 | Assistant, TU Darmstadt and University of Frankfurt | | 2005 | Dr. rer. pol., University of Frankfurt | | 2007 - 2008 | Max Weber Postdoc Fellow, European University Institute, Florence, Italy | | 2008 - 2010 | Junior Professor (W1), University of Frankfurt | | 2009 | PhD (Industrial Engineering), Politehnica University, Bucharest, Romania | | 2010 - 2014 | Professor (W2, Bonn Junior Fellow), University of Bonn | | Since 2014 | Professor (W3), University of Kiel |
|
|
My research focusses on two related aspects of time series econometrics: the analysis of persistent (fractionally integrated) time series, and exploiting the persistence to forecast time series.
Not knowing the shape of the deterministic component when testing for (co)integration poses problems; in [1], the evidence from two cointegration tests, one accounting for a deterministic linear trend in the cointegrating relations, and one without trend, is combined to obtain robust inference about the cointegration rank, which is shown to be a better procedure that pretesting or always accounting for the possible trend. The analysis of fractionally cointegrated series often presupposes knowledge of the exact degree of fractional integration. A test for the fractional integration parameter which is entirely regression based and allows for conditional and unconditional heteroskedasticity is proposed in [2].
When it comes to forecasts based on stochastic time series models, modeling the dynamics of the series is only one aspect; the other important issue is to provide the optimal forecast. Given the inherent uncertainty of forecasts, optimality is delivered from a decision-theoretic point of view by the point forecast minimizing the expected cost, or loss, caused by forecast errors. A numerical method tailored for minimization of sample counterparts of such loss functionals is put forward in [3], a method that can also be used for estimation under the relevant loss function. Moving on to forecast intervals, often used as a measure of forecast precision, it is argued in [4] that the usual,  , are not compatible with point forecasts that are optimal under loss functions different from the squared-error loss (they may not even contain the point forecast); a coherent optimality principle for forecast intervals is proposed. The issue of optimality of multivariate forecasts and of multivariate loss functions is discussed in [5].
|
|
“Approximation and aggregation in modeling and forecasting persistent time series”
Project leader, jointly with Uwe Hassler, Goethe University Frankfurt, 2011 - 2014
|
|
[ 1] Matei Demetrescu, Helmut Lütkepohl, Pentti Saikkonen
Testing for the cointegrating rank of a vector autoregressive process with uncertain deterministic trend term
Econom. J. , 12: (3): 414--435
2009
DOI: 10.1111/j.1368-423X.2009.00297.x[ 2] Matei Demetrescu, Vladimir Kuzin, Uwe Hassler
Long memory testing in the time domain
Econometric Theory , 24: (1): 176--215
2008
DOI: 10.1017/S0266466608080092[ 3] Matei Demetrescu
An extension of the Gauss-Newton algorithm for estimation under asymmetric loss
Comput. Statist. Data Anal. , 50: (2): 379--401
2006
DOI: 10.1016/j.csda.2004.08.007[ 4] Matei Demetrescu
Optimal forecast intervals under asymmetric loss
J. Forecast. , 26: (4): 227--238
2007
DOI: 10.1002/for.1019[ 7] Matei Demetrescu, Uwe Hassler
Effect of neglected deterministic seasonality on unit root tests
Statist. Papers , 48: (3): 385--402
2007
DOI: 10.1007/s00362-006-0343-6 |
|
|
|
|
|
- Master theses: 2
- Diplom theses: 1
|
|
Download Profile  |
