An Embedded Model Estimator for Non-Stationary Random Functions Using Multiple Secondary Variables

An algorithm for non-stationary spatial modelling using multiple secondary variables is developed herein, which combines geostatistics with quantile random forests to provide a new interpolation and stochastic simulation. This paper introduces the method and shows that its results are consistent and similar in nature to those applying to geostatistical modelling and to quantile random forests. The method allows for embedding of simpler interpolation techniques, such as kriging, to further condition the model. The algorithm works by estimating a conditional distribution for the target variable at each target location. The family of such distributions is called the envelope of the target variable. From this, it is possible to obtain spatial estimates, quantiles and uncertainty. An algorithm is also developed to produce conditional simulations from the envelope. As they sample from the envelope, realizations are therefore locally influenced by relative changes of importance of secondary variables, trends and variability.

Simultaneously Improve Transferability and Discriminability for Adversarial Domain Adaptation

Although adversarial domain adaptation enhances feature transferability, the feature discriminability will be degraded in the process of adversarial learning. Moreover, most domain adaptation methods only focus on distribution matching in the feature space; however, shifts in the joint distributions of input features and output labels linger in the network, and thus, the transferability is not fully exploited. In this paper, we propose a matrix rank embedding (MRE) method to enhance feature discriminability and transferability simultaneously. MRE restores a low-rank structure for data in the same class and enforces a maximum separation structure for data in different classes. In this manner, the variations within the subspace are reduced, and the separation between the subspaces is increased, resulting in improved discriminability. In addition to statistically aligning the class-conditional distribution in the feature space, MRE forces the data of the same class in different domains to exhibit an approximate low-rank structure, thereby aligning the class-conditional distribution in the label space, resulting in improved transferability. MRE is computationally efficient and can be used as a plug-and-play term for other adversarial domain adaptation networks. Comprehensive experiments demonstrate that MRE can advance state-of-the-art domain adaptation methods.

GEOPOLITICAL RISKS AND THE HIGH-FREQUENCY MOVEMENTS OF THE US TERM STRUCTURE OF INTEREST RATES

We use daily data for the period 25th November 1985 to 10th March 2020 to analyze the impact of newspapers-based measures of geopolitical risks (GPRs) on United States (US) Treasury securities by considering the level, slope and curvature factors derived from the term structure of interest rates of maturities covering 1 to 30 years. No evidence of predictability of the overall GPRs (or for threats and acts) is detected using linear causality tests. However, evidence of structural breaks and nonlinearity is provided by statistical tests performed on the linear model, which indicates that the Granger causality cannot be relied upon, as they are based on a misspecified framework. As a result, we use a data-driven approach, specifically a nonparametric causality-in-quantiles test, which is robust to misspecification due to regime changes and nonlinearity, to reconsider the predictive ability of the overall and decomposed GPRs on the three latent factors. Moreover, the zero lower bound situation, visible in our sample period, is captured by the lower quantiles, as this framework allows us to capture the entire conditional distribution of the three factors. Using this robust model, we find overwhelming evidence of causality from the GPRs, with relatively stronger effects from threats than acts, for the entire conditional distribution of the three factors, with higher impacts on medium- and long-run maturities, i.e., curvature and level factors, suggesting the predictability of the entire US term structure based on information contained in GPRs. Our results have important implications for academics, investors and policymakers.

Lossless Compression of Sensor Signals Using an Untrained Multi-Channel Recurrent Neural Predictor

The use of sensor applications has been steadily increasing, leading to an urgent need for efficient data compression techniques to facilitate the storage, transmission, and processing of digital signals generated by sensors. Unlike other sequential data such as text sequences, sensor signals have more complex statistical characteristics. Specifically, in every signal point, each bit, which corresponds to a specific precision scale, follows its own conditional distribution depending on its history and even other bits. Therefore, applying existing general-purpose data compressors usually leads to a relatively low compression ratio, since these compressors do not fully exploit such internal features. What is worse, partitioning a bit stream into groups with a preset size will sometimes break the integrity of each signal point. In this paper, we present a lossless data compressor dedicated to compressing sensor signals which is built upon a novel recurrent neural architecture named multi-channel recurrent unit (MCRU). Each channel in the proposed MCRU models a specific precision range of each signal point without breaking data integrity. During compressing and decompressing, the mirrored network will be trained on observed data; thus, no pre-training is needed. The superiority of our approach over other compressors is demonstrated experimentally on various types of sensor signals.

Yaglom limit for stochastic fluid models

In this paper we analyse the limiting conditional distribution (Yaglom limit) for stochastic fluid models (SFMs), a key class of models in the theory of matrix-analytic methods. So far, only transient and stationary analyses of SFMs have been considered in the literature. The limiting conditional distribution gives useful insights into what happens when the process has been evolving for a long time, given that its busy period has not ended yet. We derive expressions for the Yaglom limit in terms of the singularity˜$s^*$ such that the key matrix of the SFM, ${\boldsymbol{\Psi}}(s)$, is finite (exists) for all $s\geq s^*$ and infinite for $s<s^*$. We show the uniqueness of the Yaglom limit and illustrate the application of the theory with simple examples.

Quantile Regression for Panel Data and Factor Models

For nearly 25 years, advances in panel data and quantile regression were developed almost completely in parallel, with no intersection until the work by Koenker in the mid-2000s. The early theoretical work in statistics and economics raised more questions than answers, but it encouraged the development of several promising new approaches and research that offered a better understanding of the challenges and possibilities at the intersection of the literatures. Panel data quantile regression allows the estimation of effects that are heterogeneous throughout the conditional distribution of the response variable while controlling for individual and time-specific confounders. This type of heterogeneous effect is not well summarized by the average effect. For instance, the relationship between the number of students in a class and average educational achievement has been extensively investigated, but research also shows that class size affects low-achieving and high-achieving students differently. Advances in panel data include several methods and algorithms that have created opportunities for more informative and robust empirical analysis in models with subject heterogeneity and factor structure.

On a Skew t–discrete Gamma Alternative to the Normal-Poisson Model for Clustered Count Outcomes

The normal and Poisson distribution assumptions in the normal-Poisson mixed effects regression model are often too restrictive for many real count data. Several works have independently relaxed the Poisson conditional distribution assumption for counts or the normal distribution assumption for random effects. This work couples some recent advances in these two regards to develop a skew t&ndash;discrete gamma regression model in which the count outcomes have full dispersion flexibility and random effets can be skewed and heavy tailed. Inference in the model is achieved by maximum likelihood using pseudo-adaptive Gaussian quadature. The use of the proposal is demonstrated on a popular owl sibling negotiation data. It appears that, for this example, the proposed approach outperforms models based on normal random effects and the Poisson or negative binomial count distribution.

Macroeconomic and Financial Risks: A Tale of Mean and Volatility

We study the joint conditional distribution of GDP growth and corporate credit spreads using a stochastic volatility VAR. Our estimates display significant cyclical co-movement in uncertainty (the volatility implied by the conditional distributions), and risk (the probability of tail events) between the two variables. We also find that the interaction between two shocks--a main business cycle shock as in Angeletos et al. (2020) and a main financial shock--is crucial to account for the variation in uncertainty and risk, especially around crises. Our results highlight the importance of using multivariate nonlinear models to understand the determinants of uncertainty and risk.