Schedule spring 2017
Current facets (Pre-Master)
Thomas Mikosch (University of Copenhagen)
19 January 2017 **CANCELED**
The auto- and cross-distance correlation functions of a multivariate time series and their sample versions
Feuerverger (1993) and Székely, Rizzo and Bakirov (2007) introduced the notion of distance covariance/correlation as a measure of independence/dependence between two vectors of arbitrary dimension and provided limit theory for the sample versions based on an i.i.d. sequence. The main idea is to use characteristic functions to test for independence between vectors, using the standard property that the characteristic function of two independent vectors factorizes. Distance covariance is a weighted version of the squared distance between the joint characteristic function of the vectors and the product of their marginal characteristic functions. Similar ideas have been used in the literature for various purposes: goodness-of-fit tests, change point detection, testing for independence of variables, etc; see work by Meintanis, Huškova, and many others. In contrast to Székely et al. who use a weight function which is infinite on the axes, the latter authors choose probability density weights. Z. Zhou (2012) extended distance correlation to time series models for testing dependence/independence in a time series at a given lag. He assumed a “ physical dependence measure''. In our work we consider the distance covariance/correlation for general weight measures, finite or infinite on the axes or at the origin. These include the choice of Székely et al., probability and various Lévy measures. The sample versions of distance covariance/correlation are obtained by replacing the characteristic functions by their sample versions. We show consistency under ergodicity and weak convergence to an unfamiliar limit distribution of the scaled auto- and cross-distance covariance/correlation functions under strong mixing. We also study the auto-distance correlation function of the residual process of an autoregressive process. The limit theory is distinct from the corresponding theory of an i.i.d. noise process. We illustrate the theory for simulated and real data examples.
Kirill Evdokimov (Princeton University)
16 March 2017
Efficient Estimation with a Finite Number of Simulation Draws per Observation
In microeconometric applications, simulation methods such as the Method of Simulated Moments (MSM) and Indirect Inference (II) typically provide consistent and asymptotically normal estimators when a finite number of simulation draws per observation is used. However, these estimators are inefficient, unless the number of simulation draws per observation is large (theoretically, infinite). This paper argues that this inefficiency can be attributed to the standard estimators ignoring important information about the estimation problem. The paper proves that asymptotically efficient estimation is possible with as little as one simulation draw per observation, as long as the estimators make proper use of the available information. Moreover, such efficient estimators can be taken to be simple modifications of the standard MSM and II estimators with nearly no additional computational or programming burden. In practice, the possibility of using just one simulation draw per observation could significantly reduce the estimation time for models, in which evaluation at each simulation draw and parameter value is time-consuming. This in particular includes models that require numerical computation of an optimal choice, decision, or equilibrium for each simulation draw. Such models are widespread in empirical microeconomics, including industrial organization and labor economics. To establish the properties of the new estimators, the paper develops an asymptotic theory of estimation and inference in (possibly non-smooth) moment condition models with a large number of moments. This asymptotic theory covers both the extremum and quasi-Bayesian estimators.
Zhipeng Liao (UCLA)
6 April 2017
A Uniform Vuong Test for Semi/Nonparametric Models
This paper proposes a new Vuong test for the statistical comparison of semi/non-parametric models based on a general quasi-likelihood ratio criterion. An important feature of the new test is its uniformly exact asymptotic size in the overlapping nonnested case, as well as in the easier nested and strictly nonnested cases. The uniform size control is achieved without using pretesting, sample-splitting, or simulated critical values. We also show that the test has nontrivial power against all ƴn-local alternatives and against some local alternatives that converge to the null faster than ƴn. Finally, we provide a framework for conducting uniformly valid post Vuong test inference for model parameters. The finite sample performance of the uniform test and that of the post Vuong test inference procedure are illustrated in a mean-regression example by Monte Carlo.
Co-Author Xiaoxia Shi
Drew Creal (University of Chicago)
20 April 2017
Monetary Policy Uncertainty and Economic Fluctuations
We investigate the relationship between uncertainty about monetary policy and its transmission mechanism, and economic fluctuations. We propose a new term structure model where the second moments of macroeconomic variables and yields can have a first-order effect on their dynamics. The data favors a model with two unspanned volatility factors that capture uncertainty about monetary policy and the term premium. Uncertainty contributes negatively to economic activity. Two dimensions of uncertainty react in opposite directions to a shock to the real economy, and the response of inflation to uncertainty shocks vary across different historical episodes.
Liangjun Su (Singarpore Management University)
4 May 2017
Identifying Latent Grouped Structures in Nonlinear Panels
We propose a procedure to identify latent group structures in nonlinear panel data models where some regression coefficients are heterogeneous across groups but homogenous within a group and the group number and membership are unknown. To identify the group structures, we consider the order statistics for the preliminary unconstrained consistent estimates of the regression coefficients and translate the problem of classification into the problem of breaks detection. Then we extend the sequential binary segmentation algorithm of Bai (1997) for breaks detection from the time series setup to the panel data framework. We demonstrate that our method is able to identify the true latent group structures with probability approaching one and the post-classification estimators are oracally efficient. In addition, our method has the greatest advantage of easy implementation in comparison with some competitive methods in the literature, which is desirable especially for nonlinear panel data models. To improve the finite sample performance of our method, we also consider an alternative version based on the spectral decomposition of certain estimated matrix and link our group identification issue to the community detection problem in the network literature. Simulations show that our method has good finite sample performance. We apply our method to explore how individuals' portfolio choices respond to their financial status and other characteristics using the Netherlands household panel data from year 1993 to 2015, and find two latent groups.
Martin Weidner (UCL)
30 May 2017
Fixed-Effect Regressions on Network Data
This paper studies inference on fixed effects in a linear regression model estimated from network data. An important special case of our setup is the two-way regression model, which is a workhorse method in the analysis of matched data sets. Networks are typically quite sparse and it is difficult to see how the data carry information about certain parameters. We derive bounds on the variance of the fixed-effect estimator that uncover the importance of the structure of the network. These bounds depend on the smallest non-zero eigenvalue of the (normalized) Laplacian of the network and on the degree structure of the network. The Laplacian is a matrix that describes the network and its smallest non-zero eigenvalue is a measure of connectivity, with smaller values indicating less-connected networks. These bounds yield conditions for consistent estimation and convergence rates, and allow to evaluate the accuracy of first-order approximations to the variance of the fixed-effect estimator. The bounds are also used to assess the bias and variance of estimators of moments of the fixed effects.
Co author: Koen Jochmans
For more information
The Econometric Institute Seminars are supported by: