Themes of interest are centered around scalable methods for high-dimensional dynamic
econometrics; that is, high-dimensional aspects of selection,
shrinkage (toward sparsity, toward reduced rank, ...),
identification schemes for variance decompositions and impulse responses,
summarization and visualization, optimal filtering, time-varying parameters,
mixed-frequency and missing data, real-time vintage data, etc.
They include but are not limited to:
• Dynamic modeling
(e.g., VAR’s) in high dimensions (estimation, identification, parameter breaks,
parameter evolution, stochastic volatility, interaction of time-varying
parameters and stochastic volatility, ...); static and
dynamic network connectedness measurement; static and dynamic network
visualization.
• Regularization
including aspects of selection (all-subsets, partial-subsets, one-shot, ...); shrinkage (Bayesian and otherwise); distillation
(factor structure, PCA, ...); hybrids (e.g., shrinkage toward factor structure,
lasso and its variants, ...); situations with T < K.
• Volatility including
realized volatility in high dimensions (i.e., empirical quadratic variation and
covariation), microstructure noise, jumps, “realized X” for various other X,
...); assembling and constraining “vast” realized covariance matrices; tick
data and inter-trade durations; long memory and self-similarity; regime
switching and multi-fractals.
• Aspects of all of the above in real-time monitoring contexts.