135 research outputs found
Maximum Entropy Bootstrap for Time Series: The meboot R Package
No full textThe maximum entropy bootstrap is an algorithm that creates an ensemble for time series inference. Stationarity is not required and the ensemble satisfies the ergodic theorem and the central limit theorem. The meboot R package implements such algorithm. This document introduces the procedure and illustrates its scope by means of several guided applications.
GMM and OLS Estimation and Inference for New Keynesian Phillips Curve
No full textThis paper considers estimation situations where identification, endogeneity and non-spherical regression error problems are present. Instead of always using GMM despite weak instruments to solve the endogeneity, it is possible to first check whether endogeneity is serious enough to cause inconsistency in the particular problem at hand. We show how to use Maximum Entropy bootstrap (meboot) for nonstationary time series data and check `convergence in probability' and `almost sure convergence' by evaluating the proportion of sample paths straying outside error bounds as the sample size increases. The new Keynesian Phillips curve (NKPC) ordinary least squares (OLS) estimation for US data finds little endogeneity-induced inconsistency and that GMM seems to worsen it. The potential `lack of identification' problem is solved by replacing the traditional pivot which divides an estimate by its standard error by the Godambe pivot, as explained in Vinod (2008) and Vinod (2010), leading to superior confidence intervals for deep parameters of the NKPC model.Bootstrap, simulation, convergence, inflation inertia, sticky prices
If deficits are not the culprit, what determines Indian interest rates? An evaluation using the maximum entropy bootstrap method
No full textThis paper challenges two clichés that have dominated the macroeconometric debates in India. One relates to the neoclassical view that deficits are detrimental to growth, as they increase the rate of interest, and in turn displace the interest-rate-sensitive components of private investment. The second relates to the assumption of "stationarity" - which has dominated the statistical inference in time-series econometrics for a long time - as well as the emphasis on unit root-type testing, which involves detrending, or differencing, of the series to achieve stationarity in time-series econometric models. The paper examines the determinants of rates of interest in India for the periods 1980-81 and 2011-12, using the maximum entropy bootstrap (Meboot) methodology proposed in Vinod 1985 and 2004 (and developed extensively in Vinod 2006, Vinod and Lopez-de-Lacalle 2009, and Vinod 2010 and 2013). The practical appeal of Meboot is that it does not necessitate all pretests, such as structural change and unit root-type testing, which involve detrending the series to achieve stationarity, which in turn is problematic for evolutionary short time series. It also solves problems related to situations where stationarity assumptions are difficult to verify - for instance, in mixtures of I(0) and nonstationary I(d) series, where the order of integration can be different for different series. [...
A New Solution to Time Series Inference in Spurious Regression Problems
No full textPhillips (1986) provides asymptotic theory for regressions that relate nonstationary time series including those integrated of order 1, I(1). A practical implication of the literature on spurious regression is that one cannot trust the usual confidence intervals. In the absence of prior knowledge that two series are cointegrated, it is therefore recommended that after carrying out unit root tests we work with differenced or detrended series instead of original data in levels. We propose a new alternative for obtaining confidence intervals based on the Maximum Entropy bootstrap explained in Vinod and Lopez-de-Lacalle (2009). An extensive Monte Carlo simulation shows that our proposal can provide more reliable conservative confidence intervals than traditional, differencing and block bootstrap (BB) intervals.Bootstrap, simulation, confidence intervals
Reference Growth Charts for Saudi Arabian Children and Adolescents
No full textThe purpose of this study is to provide Saudi Arabian populationreference growth standards for height, weight, body mass index (BMI), headcircumference and weight for length/stature. The estimated distribution centiles are obtained by splitting the population into two separate age groups:infants, birth to 36 months and children and adolescents, age 2 to 19 years.The reference values were derived from cross-sectional data applying the LMSmethod of Cole and Green (Statistics in Medicine 1992; 11:1305–1319) usingthe lmsqreg package in R (public domain language for data analysis, 2009).The report provides an overview of how the method has been applied, morespecifically how the relevant issues concerning the construction of the growthcharts have been addressed, and is illustrated by just using the girls’ weightdata (birth to 3 years old). These issues include identifying the outliers, diagnosing the appropriate amounts of smoothing and averaging the referencestandards for the overlapping 2- to 3-year age range. The use of ANCOVAhas been introduced and illustrated as a tool for making growth standardcomparisons between different geographical regions and between genders
Statistical validation of functional form in multiple regression using R
No full textIn applied statistical research the practitioner frequently faces the problem that there neither is clear guidance from grounds of theoretical reasoning nor exists empirical (meta) evidence on the choice of functional form of a tentative regression model. Thus, parametric modeling resulting in a parametric benchmark model may easily miss important features of the data. Using recently advanced nonparametric regression methods we illustrate two powerful techniques to validate a parametric benchmark model. We discuss an empirical example using a well-known data set and provide R code snippets for the implementation of simulations and examples
Multivariate GARCH models for large-scale applications : a survey
No full textThis chapter provides a survey of various multivariate GARCH specifications that model the temporal dependence in the second moment of multivariate return series processes. The survey is focused on feasible multivariate GARCH models for large-scale applications, as well as on recent contributions in outlier-robust MGARCH analysis and the use of high-frequency returns or the score for covariance modeling. We discuss their likelihood-based estimation and application to forecasting and simulation with software implementations in the R-programming language.</p
Hypothesis testing, specification testing and model selection based on the MCMC output using R
No full textKernel Regression Coefficients for Practical Significance
No full textQuantitative researchers often use Student’s t-test (and its p-values) to claim that a particular regressor is important (statistically significantly) for explaining the variation in a response variable. A study is subject to the p-hacking problem when its author relies too much on formal statistical significance while ignoring the size of what is at stake. We suggest reporting estimates using nonlinear kernel regressions and the standardization of all variables to avoid p-hacking. We are filling an essential gap in the literature because p-hacking-related papers do not even mention kernel regressions or standardization. Although our methods have general applicability in all sciences, our illustrations refer to risk management for a cross-section of firms and financial management in macroeconomic time series. We estimate nonlinear, nonparametric kernel regressions for both examples to illustrate the computation of scale-free generalized partial correlation coefficients (GPCCs). We suggest supplementing the usual p-values by “practical significance” revealed by scale-free GPCCs. We show that GPCCs also yield new pseudo regression coefficients to measure each regressor’s relative (nonlinear) contribution in a kernel regression
- …
