scholarly journals Local Projection Inference Is Simpler and More Robust Than You Think

Econometrica ◽  
2021 ◽  
Vol 89 (4) ◽  
pp. 1789-1823
Author(s):  
José Luis Montiel Olea ◽  
Mikkel Plagborg-Møller
Keyword(s):  
Serial Correlation ◽  
Standard Errors ◽  
Impulse Responses ◽  
Local Projection ◽  
Regression Residuals ◽  
Wide Range ◽  
Persistent Data ◽  
Future Outcomes ◽  
Local Projections ◽  
Linear Regressions

Applied macroeconomists often compute confidence intervals for impulse responses using local projections, that is, direct linear regressions of future outcomes on current covariates. This paper proves that local projection inference robustly handles two issues that commonly arise in applications: highly persistent data and the estimation of impulse responses at long horizons. We consider local projections that control for lags of the variables in the regression. We show that lag‐augmented local projections with normal critical values are asymptotically valid uniformly over (i) both stationary and non‐stationary data, and also over (ii) a wide range of response horizons. Moreover, lag augmentation obviates the need to correct standard errors for serial correlation in the regression residuals. Hence, local projection inference is arguably both simpler than previously thought and more robust than standard autoregressive inference, whose validity is known to depend sensitively on the persistence of the data and on the length of the horizon.

The analysis of the results of a rotation experiment on the use of straw and fertilizers

1953 ◽  
Vol 43 (1) ◽  
pp. 77-88 ◽  
Author(s):  
H. D. Patterson
Keyword(s):  
Present Experiment ◽  
Treatment Effects ◽  
Standard Errors ◽  
Methods Of Analysis ◽  
Seasonal Factors ◽  
Long Term Experiments ◽  
The Mean ◽  
Linear Regressions

An experiment, designed to test different ways of using straw with fertilizers, and involving a three course rotation of crops, was carried out at Rothamsted between 1933 and 1951. The methods of analysis developed for this experiment are described in the present paper and demonstrated using yields of potatoes.Treatment effects of interest are given by the mean yields over all years and the linear regressions of yield on time. These estimates are straightforward but the evaluation of their errors is complicated by the existence of correlations due to the recurrence of treatments on the same plots. Further complications are introduced when, as frequently happens in long-term experiments, treatment effects show real variation from year to year. A method is given for estimating standard errors which include a contribution from this variation.The various relationships between yields and the uncontrolled seasonal factors can also be examined; in the present experiment there is some indication that the effects of treatments on yields of potatoes are influenced by the dates of planting.In other circumstances the analysis requires modifications, some of which are briefly considered.

Using Placebo Laws to Test "More Guns, Less Crime"

2004 ◽  
Vol 4 (1) ◽  
Author(s):  
Eric Helland ◽  
Alexander Tabarrok
Keyword(s):  
Public Policy ◽  
Standard Error ◽  
Serial Correlation ◽  
Error Function ◽  
Standard Errors ◽  
Grouped Data ◽  
Natural Experiments ◽  
Statistical Assumptions ◽  
Gun Laws ◽  
Empirical Standard

Abstract We reexamine Mustard and Lott’s controversial study on the effect of “shall-issue” gun laws on crime using an empirical standard error function randomly generated from “placebo” laws. We find that the effect of shall-issue laws on crime is much less well-estimated than the Mustard and Lott (1997) and Lott (2000) results suggest. We also find, however, that the cross equation restrictions implied by the Lott-Mustard theory are supported. A boomlet has occurred in recent years in the use of quasi-natural experiments to answer important questions of public policy. The intuitive power of this approach, however, has sometimes diverted attention from the statistical assumptions that must be made, particularly regarding standard errors. Failing to take into account serial correlation and grouped data can dramatically reduce standard errors suggesting greater certainty in effects than is actually the case. We find that the placebo law technique (Bertrand, Duflo and Mullainathan 2002) is a useful addition to the econometrician’s toolkit.

Estimation and Inference of Impulse Responses by Local Projections

2005 ◽  
Vol 95 (1) ◽  
pp. 161-182 ◽  
Author(s):  
Òscar Jordà
Keyword(s):  
New Keynesian ◽  
Closed Economy ◽  
Impulse Responses ◽  
Vector Autoregressions ◽  
Central Idea ◽  
Keynesian Model ◽  
Highly Nonlinear ◽  
Regression Techniques ◽  
Local Projections ◽  
Simple Regression

This paper introduces methods to compute impulse responses without specification and estimation of the underlying multivariate dynamic system. The central idea consists in estimating local projections at each period of interest rather than extrapolating into increasingly distant horizons from a given model, as it is done with vector autoregressions (VAR). The advantages of local projections are numerous: (1) they can be estimated by simple regression techniques with standard regression packages; (2) they are more robust to misspecification; (3) joint or point-wise analytic inference is simple; and (4) they easily accommodate experimentation with highly nonlinear and flexible specifications that may be impractical in a multivariate context. Therefore, these methods are a natural alternative to estimating impulse responses from VARs. Monte Carlo evidence and an application to a simple, closed-economy, new-Keynesian model clarify these numerous advantages.

scholarly journals Motor and non-motor sequence prediction is equally affected in children with Developmental Coordination Disorder

2020 ◽  
Author(s):  
Bertram Opitz ◽  
Daniel Brady ◽  
Hayley C. Leonard
Keyword(s):  
Developmental Coordination Disorder ◽  
Motor Pattern ◽  
Motor Task ◽  
Forward Modelling ◽  
Motor Sequence ◽  
Cognitive Domains ◽  
Internal Forward Model ◽  
Wide Range ◽  
Future Outcomes ◽  
Coordination Disorder

AbstractChildren with Developmental Coordination Disorder (DCD) are diagnosed based on motor difficulties. However, they also exhibit difficulties in several other cognitive domains, including visuospatial processing, executive functioning and attention. One account of the difficulties seen in DCD proposes an impairment in internal forward modelling, i.e., the ability to (i) detect regularities of a repetitive perceptual or motor pattern, (ii) predict future outcomes of motor actions, and (iii) adapt behaviour accordingly. Using electroencephalographic recordings, the present study aimed to delineate these different aspects of internal forward modelling across several domains. To this end, 24 children with DCD and 23 typically-developing children (aged 7-10 years) completed a serial prediction task in the visual, temporal, spatial and motor domains. This task required them to learn short sequences and to indicate whether a sequence was disrupted towards its end. Analyses revealed that, across all domains, children with DCD showed poorer discrimination between intact and disrupted sequences, accompanied by a delayed late parietal positivity elicited by disrupted sequences. These results indicate an impairment in explicit sequence discrimination in DCD across motor and cognitive domains. However, there is no evidence for an impairment in implicit performance on the motor task in DCD. These results suggest an impairment of the updating of an internal forward model in DCD resulting in a blurred representation of that model and consequently in a reduced ability to detect regularities in the environment (e.g., sequences). Such a detailed understanding of internal forward modelling in DCD could help to explain the wide range of co-occurring difficulties experienced by those with a diagnosis of DCD.

scholarly journals Local Projections and VARs Estimate the Same Impulse Responses

Econometrica ◽  
2021 ◽  
Vol 89 (2) ◽  
pp. 955-980
Author(s):  
Mikkel Plagborg-Møller ◽  
Christian K. Wolf
Keyword(s):  
Structural Identification ◽  
Finite Sample ◽  
Impulse Responses ◽  
Vector Autoregressions ◽  
Reduction Techniques ◽  
Long Run ◽  
Finite Sample Properties ◽  
Short Run ◽  
Local Projections ◽  
Dimension Reduction Techniques

We prove that local projections (LPs) and Vector Autoregressions (VARs) estimate the same impulse responses. This nonparametric result only requires unrestricted lag structures. We discuss several implications: (i) LP and VAR estimators are not conceptually separate procedures; instead, they are simply two dimension reduction techniques with common estimand but different finite‐sample properties. (ii) VAR‐based structural identification—including short‐run, long‐run, or sign restrictions—can equivalently be performed using LPs, and vice versa. (iii) Structural estimation with an instrument (proxy) can be carried out by ordering the instrument first in a recursive VAR, even under noninvertibility. (iv) Linear VARs are as robust to nonlinearities as linear LPs.

scholarly journals Semi-automated simultaneous predictor selection for regression-SARIMA models

2020 ◽  
Vol 30 (6) ◽  
pp. 1759-1778
Author(s):  
Aaron P. Lowther ◽  
Paul Fearnhead ◽  
Matthew A. Nunes ◽  
Kjeld Jensen
Keyword(s):  
Linear Models ◽  
Model Fitting ◽  
Selection Procedure ◽  
Predictive Performance ◽  
Mixed Integer ◽  
Linear Regression Models ◽  
Regression Residuals ◽  
Wide Range ◽  
Telecommunications Network ◽  
Integral Role

Abstract Deciding which predictors to use plays an integral role in deriving statistical models in a wide range of applications. Motivated by the challenges of predicting events across a telecommunications network, we propose a semi-automated, joint model-fitting and predictor selection procedure for linear regression models. Our approach can model and account for serial correlation in the regression residuals, produces sparse and interpretable models and can be used to jointly select models for a group of related responses. This is achieved through fitting linear models under constraints on the number of nonzero coefficients using a generalisation of a recently developed mixed integer quadratic optimisation approach. The resultant models from our approach achieve better predictive performance on the motivating telecommunications data than methods currently used by industry.

Comparison of methods for prediction of rumen fermentation patterns from diet composition

2001 ◽  
Vol 2001 ◽  
pp. 27-27
Author(s):  
V.E. Brown ◽  
R.E. Agnew ◽  
D.J. Kilpatrick
Keyword(s):  
Diet Composition ◽  
Near Infrared ◽  
Chemical Constituents ◽  
Rumen Fermentation ◽  
Prediction System ◽  
Multiple Linear Regressions ◽  
Near Infrared Reflectance ◽  
Wet Chemistry ◽  
Wide Range ◽  
Linear Regressions

Previous attempts (Offer & Percival, 1998) have been made to develop a prediction system for rumen fermentation patterns from stepwise multiple linear regressions of the chemical constituents of the diet. These authors have also made comparisons between equations derived from diet wet chemistry and those developed from near infrared reflectance spectroscopy (NIRS). However, the potential of NIRS to predict the dynamics of rumen fermentation has not fully been explored using a wide range of forage treatments. Therefore the objective of this experiment was to develop equations from the chemical composition of the diet to predict rumen fermentation patterns and compare these with equations developed from undried and dried NIRS scans of the diets.

Bias in common catch-curve methods applied to age frequency data from fish surveys

2019 ◽  
Vol 76 (7) ◽  
pp. 2090-2101
Author(s):  
Gary A Nelson
Keyword(s):  
Linear Regression ◽  
Mixed Model ◽  
Linear Mixed Model ◽  
Standard Errors ◽  
Cluster Sampling ◽  
Probability Models ◽  
Generalized Poisson ◽  
Random Intercept ◽  
Multi Stage ◽  
Wide Range

Abstract Catch curve analysis is often used in data-limited fisheries stock assessments to estimate total instantaneous mortality (Z). There are now six catch-curve methods available in the literature: the Chapman–Robson, linear regression, weighted linear regression, Heincke, generalized Poisson linear, and random-intercept Poisson linear mixed model. An assumption shared among the underyling probability models of these estimators is that fish collected for ageing are sampled from the population by simple random sampling. This type of sampling is nearly impossible in fisheries research because populations are sampled in surveys that use gears that capture individuals in clusters and often fish for ageing are selected from multi-stage sampling. In this study, I explored the effects of multi-stage cluster sampling on the bias of the estimates of Z and their associated standard errors. I found that the generalized Poisson linear model and the Chapman–Robson estimators were the least biased, whereas the random-intercept Poisson linear mixed model was the most biased under a wide range of simulation scenarios that included different levels of recruitment variation, intra-cluster correlation, sample sizes, and methods used to generate age frequencies. Standard errors of all estimators were under-estimated in almost all cases and should not be used in statistical comparisons.

scholarly journals Efficient Standard Errors in Item Response Theory Models for Short Tests

2019 ◽  
Vol 80 (3) ◽  
pp. 461-475
Author(s):  
Lianne Ippel ◽  
David Magis
Keyword(s):  
Item Response Theory ◽  
Item Response ◽  
Standard Errors ◽  
Response Theory ◽  
Asymptotic Standard Error ◽  
Irt Models ◽  
Global Comparison ◽  
Wide Range ◽  
Item Response Theory Models ◽  
Theory Framework

In dichotomous item response theory (IRT) framework, the asymptotic standard error (ASE) is the most common statistic to evaluate the precision of various ability estimators. Easy-to-use ASE formulas are readily available; however, the accuracy of some of these formulas was recently questioned and new ASE formulas were derived from a general asymptotic theory framework. Furthermore, exact standard errors were suggested to better evaluate the precision of ability estimators, especially with short tests for which the asymptotic framework is invalid. Unfortunately, the accuracy of exact standard errors was assessed so far only in a very limiting setting. The purpose of this article is to perform a global comparison of exact versus (classical and new formulations of) asymptotic standard errors, for a wide range of usual IRT ability estimators, IRT models, and with short tests. Results indicate that exact standard errors globally outperform the ASE versions in terms of reduced bias and root mean square error, while the new ASE formulas are also globally less biased than their classical counterparts. Further discussion about the usefulness and practical computation of exact standard errors are outlined.

scholarly journals Usporedna studija predviđanja koeficijenta molekularne difuzije za polarni i nepolarni binarni plin pomoću neuronskih mreža i višestrukih linearnih regresija

2019 ◽  
Vol 68 (11-12) ◽  
pp. 573-582 ◽  
Author(s):  
Naima Melzi ◽  
Hamid Zentou ◽  
Maamar Laidi ◽  
Salah Hanini ◽  
Yamina Ammi ◽  
...  
Keyword(s):  
Neural Network ◽  
Artificial Neural Network ◽  
Molecular Diffusion ◽  
Mean Squared Error ◽  
External Validation ◽  
Multiple Linear Regressions ◽  
Mean Squared Errors ◽  
Wide Range ◽  
Artificial Neural ◽  
Linear Regressions

In the current study, an artificial neural network (ANN) and multiple linear regressions (MLR) have been used to develop predictive models for the estimation of molecular diffusion coefficients of 1252 polar and non-polar binary gases at multiple pressures over a wide range of temperatures and substances. The quality and reliability of each method were estimated in terms of the correlation coefficient (R), mean squared errors (MSE), root mean squared error (RMSE), and in terms of external validation coefficients (Q2ext). The comparison between the artificial neural network (ANN) and the multiple linear regressions (MLR) revealed that the neural network models showed a good predicting ability with lower errors (the roots of the mean squared errors in the total database were 0.1400 for ANN1 and 0.1300 for ANN2), and (root mean squared errors in the total databases were 0.5172 for MLR1 and 0.5000 for MLR2).

Sign in / Sign up

Export Citation Format

Share Document