Search

Dynamic stochastic general equilibrium (DSGE) models have been an ubiquitous, and controversial, part of macroeconomics for decades. In this paper, we approach DSGEs purely as statstical models. We do this by applying two common model validation checks to the canonical Smets and Wouters 2007 DSGE: (1) we simulate the model and see how well it can be estimated from its own simulation output, and (2) we see how well it can seem to fit nonsense data. We find that (1) even with centuries' worth of data, the model remains poorly estimated, and (2) when we swap series at random, so that (e.g.) what the model gets as the inflation rate is really hours worked, what it gets as hours worked is really investment, etc., the fit is often only slightly impaired, and in a large percentage of cases actually improves (even out of sample). Taken together, these findings cast serious doubt on the meaningfulness of parameter estimates for this DSGE, and on whether this specification represents anything structural about the economy. Constructively, our approaches can be used for model validation by anyone working with macroeconomic time series., Comment: 36 pages, 21 figures, 7 tables

Trend filtering is a modern approach to nonparametric regression that is more adaptive to local smoothness than splines or similar basis procedures. Existing analyses of trend filtering focus on estimating a function corrupted by homoskedastic Gaussian noise, but our work extends this technique to general exponential family distributions. This extension is motivated by the need to study massive, gridded climate data derived from polar-orbiting satellites. We present algorithms tailored to large problems, theoretical results for general exponential family likelihoods, and principled methods for tuning parameter selection without excess computation., Comment: 53 pages; 6 figures; 3 tables

The sparse group lasso is a high-dimensional regression technique that is useful for problems whose predictors have a naturally grouped structure and where sparsity is encouraged at both the group and individual predictor level. In this paper we discuss a new R package for computing such regularized models. The intention is to provide highly optimized solution routines enabling analysis of very large datasets, especially in the context of sparse design matrices., Comment: 18 pages, 9 figures, 1 table

Forecasting methodologies have always attracted a lot of attention and have become an especially hot topic since the beginning of the COVID-19 pandemic. In this paper we consider the problem of multi-period forecasting that aims to predict several horizons at once. We propose a novel approach that forces the prediction to be "smooth" across horizons and apply it to two tasks: point estimation via regression and interval prediction via quantile regression. This methodology was developed for real-time distributed COVID-19 forecasting. We illustrate the proposed technique with the CovidCast dataset as well as a small simulation example.

Methods for global measurement of transcript abundance such as microarrays and RNA-Seq generate datasets in which the number of measured features far exceeds the number of observations. Extracting biologically meaningful and experimentally tractable insights from such data therefore requires high-dimensional prediction. Existing sparse linear approaches to this challenge have been stunningly successful, but some important issues remain. These methods can fail to select the correct features, predict poorly relative to non-sparse alternatives, or ignore any unknown grouping structures for the features. We propose a method called SuffPCR that yields improved predictions in high-dimensional tasks including regression and classification, especially in the typical context of omics with correlated features. SuffPCR first estimates sparse principal components and then estimates a linear model on the recovered subspace. Because the estimated subspace is sparse in the features, the resulting predictions will depend on only a small subset of genes. SuffPCR works well on a variety of simulated and experimental transcriptomic data, performing nearly optimally when the model assumptions are satisfied. We also demonstrate near-optimal theoretical guarantees., Comment: 28 pages, 9 figures, 10 tables

For concertgoers, musical interpretation is the most important factor in determining whether or not we enjoy a classical performance. Every performance includes mistakes -- intonation issues, a lost note, an unpleasant sound -- but these are all easily forgotten (or unnoticed) when a performer engages her audience, imbuing a piece with novel emotional content beyond the vague instructions inscribed on the printed page. In this research, we use data from the CHARM Mazurka Project -- forty-six professional recordings of Chopin's Mazurka Op. 68 No. 3 by consummate artists -- with the goal of elucidating musically interpretable performance decisions. We focus specifically on each performer's use musical tempo by examining the inter-onset intervals of the note attacks in the recording. To explain these tempo decisions, we develop a switching state space model and estimate it by maximum likelihood combined with prior information gained from music theory and performance practice. We use the estimated parameters to quantitatively describe individual performance decisions and compare recordings. These comparisons suggest methods for informing music instruction, discovering listening preferences, and analyzing performances., Comment: 50 pages, 30 figures, 9 tables

Motivation: The discovery of relationships between gene expression measurements and phenotypic responses is hampered by both computational and statistical impediments. Conventional statistical methods are less than ideal because they either fail to select relevant genes, predict poorly, ignore the unknown interaction structure between genes, or are computationally intractable. Thus, the creation of new methods which can handle many expression measurements on relatively small numbers of patients while also uncovering gene-gene relationships and predicting well is desirable. Results: We develop a new technique for using the marginal relationship between gene expression measurements and patient survival outcomes to identify a small subset of genes which appear highly relevant for predicting survival, produce a low-dimensional embedding based on this small subset, and amplify this embedding with information from the remaining genes. We motivate our methodology by using gene expression measurements to predict survival time for patients with diffuse large B-cell lymphoma, illustrate the behavior of our methodology on carefully constructed synthetic examples, and test it on a number of other gene expression datasets. Our technique is computationally tractable, generally outperforms other methods, is extensible to other phenotypes, and also identifies different genes (relative to existing methods) for possible future study. Key words: regression; principal components; matrix sketching; preconditioning Availability: All of the code and data are available at https://github.com/dajmcdon/aimer/., Comment: 22 pages, 5 figures

Trends in terrestrial temperature variability are perhaps more relevant for species viability than trends in mean temperature. In this paper, we develop methodology for estimating such trends using multi-resolution climate data from polar orbiting weather satellites. We derive two novel algorithms for computation that are tailored for dense, gridded observations over both space and time. We evaluate our methods with a simulation that mimics these data's features and on a large, publicly available, global temperature dataset with the eventual goal of tracking trends in cloud reflectance temperature variability., Comment: Published in AAAI-19

Modern applications require methods that are computationally feasible on large datasets but also preserve statistical efficiency. Frequently, these two concerns are seen as contradictory: approximation methods that enable computation are assumed to degrade statistical performance relative to exact methods. In applied mathematics, where much of the current theoretical work on approximation resides, the inputs are considered to be observed exactly. The prevailing philosophy is that while the exact problem is, regrettably, unsolvable, any approximation should be as small as possible. However, from a statistical perspective, an approximate or regularized solution may be preferable to the exact one. Regularization formalizes a trade-off between fidelity to the data and adherence to prior knowledge about the data-generating process such as smoothness or sparsity. The resulting estimator tends to be more useful, interpretable, and suitable as an input to other methods. In this paper, we propose new methodology for estimation and prediction under a linear model borrowing insights from the approximation literature. We explore these procedures from a statistical perspective and find that in many cases they improve both computational and statistical performance., Comment: 39 pages, 9 figures

This paper presents minimax rates for density estimation when the data dimension $d$ is allowed to grow with the number of observations $n$ rather than remaining fixed as in previous analyses. We prove a non-asymptotic lower bound which gives the worst-case rate over standard classes of smooth densities, and we show that kernel density estimators achieve this rate. We also give oracle choices for the bandwidth and derive the fastest rate $d$ can grow with $n$ to maintain estimation consistency., Comment: 10 pages; accepted at The 20th International Conference on Artificial Intelligence and Statistics (AISTATS)

High-dimensional predictive models, those with more measurements than observations, require regularization to be well defined, perform well empirically, and possess theoretical guarantees. The amount of regularization, often determined by tuning parameters, is integral to achieving good performance. One can choose the tuning parameter in a variety of ways, such as through resampling methods or generalized information criteria. However, the theory supporting many regularized procedures relies on an estimate for the variance parameter, which is complicated in high dimensions. We develop a suite of information criteria for choosing the tuning parameter in lasso regression by leveraging the literature on high-dimensional variance estimation. We derive intuition showing that existing information-theoretic approaches work poorly in this setting. We compare our risk estimators to existing methods with an extensive simulation and derive some theoretical justification. We find that our new estimators perform well across a wide range of simulation conditions and evaluation criteria., Comment: 64 pages, 11 figures

In this paper we analyze approximate methods for undertaking a principal components analysis (PCA) on large data sets. PCA is a classical dimension reduction method that involves the projection of the data onto the subspace spanned by the leading eigenvectors of the covariance matrix. This projection can be used either for exploratory purposes or as an input for further analysis, e.g. regression. If the data have billions of entries or more, the computational and storage requirements for saving and manipulating the design matrix in fast memory is prohibitive. Recently, the Nystr\"om and column-sampling methods have appeared in the numerical linear algebra community for the randomized approximation of the singular value decomposition of large matrices. However, their utility for statistical applications remains unclear. We compare these approximations theoretically by bounding the distance between the induced subspaces and the desired, but computationally infeasible, PCA subspace. Additionally we show empirically, through simulations and a real data example involving a corpus of emails, the trade-off of approximation accuracy and computational complexity., Comment: 20 pages

The lasso and related sparsity inducing algorithms have been the target of substantial theoretical and applied research. Correspondingly, many results are known about their behavior for a fixed or optimally chosen tuning parameter specified up to unknown constants. In practice, however, this oracle tuning parameter is inaccessible so one must use the data to select one. Common statistical practice is to use a variant of cross-validation for this task. However, little is known about the theoretical properties of the resulting predictions with such data-dependent methods. We consider the high-dimensional setting with random design wherein the number of predictors $p$ grows with the number of observations $n$. Under typical assumptions on the data generating process, similar to those in the literature, we recover oracle rates up to a log factor when choosing the tuning parameter with cross-validation. Under weaker conditions, when the true model is not necessarily linear, we show that the lasso remains risk consistent relative to its linear oracle. We also generalize these results to the group lasso and square-root lasso and investigate the predictive and model selection performance of cross-validation via simulation., Comment: 25 pages, 3 figures

We derive generalization error bounds for traditional time-series forecasting models. Our results hold for many standard forecasting tools including autoregressive models, moving average models, and, more generally, linear state-space models. These non-asymptotic bounds need only weak assumptions on the data-generating process, yet allow forecasters to select among competing models and to guarantee, with high probability, that their chosen model will perform well. We motivate our techniques with and apply them to standard economic and financial forecasting tools---a GARCH model for predicting equity volatility and a dynamic stochastic general equilibrium model (DSGE), the standard tool in macroeconomic forecasting. We demonstrate in particular how our techniques can aid forecasters and policy makers in choosing models which behave well under uncertainty and mis-specification., Comment: 34 pages, 3 figures

The lasso procedure is ubiquitous in the statistical and signal processing literature, and as such, is the target of substantial theoretical and applied research. While much of this research focuses on the desirable properties that lasso possesses---predictive risk consistency, sign consistency, correct model selection---all of it has assumes that the tuning parameter is chosen in an oracle fashion. Yet, this is impossible in practice. Instead, data analysts must use the data twice, once to choose the tuning parameter and again to estimate the model. But only heuristics have ever justified such a procedure. To this end, we give the first definitive answer about the risk consistency of lasso when the smoothing parameter is chosen via cross-validation. We show that under some restrictions on the design matrix, the lasso estimator is still risk consistent with an empirically chosen tuning parameter., Comment: 15 pages, 0 figures

Vapnik-Chervonenkis (VC) dimension is a fundamental measure of the generalization capacity of learning algorithms. However, apart from a few special cases, it is hard or impossible to calculate analytically. Vapnik et al. [10] proposed a technique for estimating the VC dimension empirically. While their approach behaves well in simulations, it could not be used to bound the generalization risk of classifiers, because there were no bounds for the estimation error of the VC dimension itself. We rectify this omission, providing high probability concentration results for the proposed estimator and deriving corresponding generalization bounds., Comment: 11 pages

The literature on statistical learning for time series often assumes asymptotic independence or "mixing" of the data-generating process. These mixing assumptions are never tested, nor are there methods for estimating mixing coefficients from data. Additionally, for many common classes of processes (Markov processes, ARMA processes, etc.) general functional forms for various mixing rates are known, but not specific coefficients. We present the first estimator for beta-mixing coefficients based on a single stationary sample path and show that it is risk consistent. Since mixing rates depend on infinite-dimensional dependence, we use a Markov approximation based on only a finite memory length $d$. We present convergence rates for the Markov approximation and show that as $d\rightarrow\infty$, the Markov approximation converges to the true mixing coefficient. Our estimator is constructed using $d$-dimensional histogram density estimates. Allowing asymptotics in the bandwidth as well as the dimension, we prove $L^1$ concentration for the histogram as an intermediate step. Simulations wherein the mixing rates are calculable and a real-data example demonstrate our methodology., Comment: 30 pages, 8 figures. Longer version of arXiv:1103.0941 [stat.ML]

In many areas of machine learning, it becomes necessary to find the eigenvector decompositions of large matrices. We discuss two methods for reducing the computational burden of spectral decompositions: the more venerable Nystom extension and a newly introduced algorithm based on random projections. Previous work has centered on the ability to reconstruct the original matrix. We argue that a more interesting and relevant comparison is their relative performance in clustering and classification tasks using the approximate eigenvectors as features. We demonstrate that performance is task specific and depends on the rank of the approximation., Comment: 11 pages, 4 figures

We show how to control the generalization error of time series models wherein past values of the outcome are used to predict future values. The results are based on a generalization of standard i.i.d. concentration inequalities to dependent data without the mixing assumptions common in the time series setting. Our proof and the result are simpler than previous analyses with dependent data or stochastic adversaries which use sequential Rademacher complexities rather than the expected Rademacher complexity for i.i.d. processes. We also derive empirical Rademacher results without mixing assumptions resulting in fully calculable upper bounds., Comment: 15 pages, 1 figure

We derive generalization error bounds for stationary univariate autoregressive (AR) models. We show that imposing stationarity is enough to control the Gaussian complexity without further regularization. This lets us use structural risk minimization for model selection. We demonstrate our methods by predicting interest rate movements., Comment: 10 pages, 3 figures. CMU Statistics Technical Report

The literature on statistical learning for time series assumes the asymptotic independence or ``mixing' of the data-generating process. These mixing assumptions are never tested, nor are there methods for estimating mixing rates from data. We give an estimator for the $\beta$-mixing rate based on a single stationary sample path and show it is $L_1$-risk consistent., Comment: 9 pages, accepted by AIStats. CMU Statistics Technical Report