A Note on the Bandwidth Choice When the Null Hypothesis is Semiparametric (original) (raw)
Related papers
An Adaptive Specification Test For Semiparametric Models
SSRN Electronic Journal, 2007
This paper introduces a new test of a semiparametric model of a conditional density function against a fully nonparametric alternative. This test is motivated by the fact that many important econometric models need to be estimated through maximum likelihood type procedures, e.g. semiparametric limited dependent variable models. This specification is also important for prediction purposes. Our test statistic combines the methodology of goodness of fit tests and nonparametric methods and the specific difficulty we focus here comes from the fact that we consider a semiparametric null hypothesis and the test statistic depends on a bandwidth parameter that needs to be estimated from the data. In order to handle the previous difficulties we introduce a data adaptive testing procedure that enables us to select the bandwidth parameter in such a way that it maximizes the power of the test. It is also shown that this procedure handles properly the bias problem generated by the introduction of a semiparametric model under the null. The distribution of the standarized test statistic is approximated under the null by both bootstrap and subsampling methods and its power is studied against local alternatives to the null hypothesis. We discuss practical issues for the application statistics and illustrate in an intensive monte carlo study both the feasibility and the performance of the procedure in finite samples of moderate size. 1
Testing for additivity in nonparametric regression
1995
Additive models are one means of assuaging the curse of dimensionality when nonparametric smoothing methods are used to estimate multivariable regression functions. It is important to have methods for testing the fit of such models, especially in high dimensions where visual assessment of fit becomes difficult. New tests of additivity are proposed in this paper that derive from Fourier series estimators with data-driven smoothing parameters. Other tests related to the classical Tukey test for additivity are also considered. While the new tests are consistent against essentially any ''smooth'' alternative to additivity, the Tukey-type tests are found to be inconsistent in certain situations. Asymptotic power of both varieties of tests is studied under local alternatives that tend toward additivity at a parametric rate, and small-sample power comparisons are carried out by means of a simulation study.
Estimation and model specification testing in nonparametric and semiparametric econometric models
Mpra Paper, 2003
This paper considers two classes of semiparametric nonlinear regression models, in which nonlinear components are introduced to reflect the nonlinear fluctuation in the mean. A general estimation and testing procedure for nonparametric time series regression under the α-mixing condition is introduced. Several test statistics for testing nonparametric significance, linearity and additivity in nonparametric and semiparametric time series econometric models are then constructed. The proposed test statistics are shown to have asymptotic normal distributions under their respective null hypotheses. Moreover, the proposed testing procedures are illustrated by several simulated examples. In addition, one of the proposed testing procedures is applied to a continuoustime model and implemented through a set of the US Federal interest rate data. Our research suggests that it is unreasonable to assume the linearity in the drift for the given data as required by some existing studies.
Journal of Econometrics, 2003
This paper considers the problem of consistent model specification tests using series estimation methods. The null models we consider in this paper all contain some nonparametric components. A leading case we consider is to test for an additive partially linear model. The null distribution of the test statistic is derived using a central limit theorem for Hilbert valued random arrays. The test statistic is shown to be able to detect local alternatives that approach the null models at the order of O p (n −1/2). We suggest to use the wild bootstrap method to approximate the critical values of the test. A small Monte Carlo simulation is reported to examine the finite sample performance of the proposed test. We also show that the proposed test can be easily modified to obtain series-based consistent tests for other semiparametric/nonparametric models.
A comparison of different nonparametric methods for inference on additive models
Journal of Nonparametric Statistics, 2005
In this article we highlight the main differences of available methods for the analysis of regression functions that are probably additive separable. We first discuss definition and interpretation of the most common estimators in practice. This is done by explaining the different ideas of modeling behind each estimator as well as what the procedures are doing to the data. Computational aspects are mentioned explicitly. The illustrated discussion concludes with a simulation study on the mean squared error for different marginal integration approaches. Next, various test statistics for checking additive separability are introduced and accomplished with asymptotic theory. Based on the asymptotic results under hypothesis as well as under the alternative of non additivity we compare the tests in a brief discussion. For the various statistics, different smoothing and bootstrap methods we perform a detailed simulation study. A main focus in the reported results is directed on the (non-) reliability of the methods when the covariates are strongly correlated among themselves. Again, a further point are the computational aspects. We found that the most striking differences lie in the different pre-smoothers that are used, but less in the different constructions of test statistics. Moreover, although some of the observed differences are strong, they surprisingly can not be revealed by asymptotic theory. 1
Bootstrap inference in semiparametric generalized additive models
2004
Here, G is a known link, a, J3 are unknown parameters, and ml, ... , I1ld are unknown (smooth) functions of possibly higher dimensional covariates T}, ... , T d. Estimates of m}, ... , 11ld, a and J3 are presented and asymptotic distribution theory for both the nonparametric and the parametric part is given. The main focus is the application of bootstrap methods. It is shown that bootstrap can be used for bias correction, hypothesis testing (e.g. component-wise analysis) and the construction of uniform confidence bands. Various bootstrap tests for model specification and parametrization are given, in particular for testing additivity and link function specification. The practical performance of our methods is illustrated in simulations and in an application to East-West German migration.
Specification Testing When the Null is Nonparametric or Semiparametric
Econometric Theory, 2014
This paper discusses the problem of testing misspecifications in semiparametric regression models for a large family of econometric models under rather general conditions. We focus on two main issues that typically arise in econometrics. First, many econometric models are estimated through maximum likelihood or pseudo-ML methods like, for example, limited dependent variable or gravity models. Second, often one might not want to fully specify the null hypothesis. Instead, one would rather impose some structure like separability or monotonicity. In order to address these points we introduce an adaptive omnibus test. Special emphasis is given to practical issues like adaptive bandwidth choice, general but simple requirements on the estimates, and finite sample performance, including the resampling approximations.
Validation tests for semi-parametric models
Journal of Statistical Computation and Simulation, 2013
Tests are proposed for validation of the hypothesis that a partial linear regression model adequately describes the structure of a given data set. The test statistics are formulated following the approach of Fourier-type conditional expectations first suggested by Bierens (1982). The proposed procedures are computationally convenient, and under fairly mild conditions lead to consistent tests. Corresponding bootstrap versions are compared with alternative procedures for a wide selection of different estimators of the underlying partial linear model.
Nonparametric Regression and Causality Testing:A Monte-Carlo Study
Scottish Journal of Political Economy, 1998
In this paper we propose a new procedure for causality testing between two univariate time-series using nonparametric additive models (see also Bell et al., 1996, 1994). We argue that the major advantage of our proposed method is that it can be applied if the underlying data generation process (DGP) is either linear or nonlinear. 1 In contrast, the predominate causality testing paradigm-based on the work of Wiener (1956) and Granger (1963, 1969)-assumes that the underlying DGP is linear and the stochastic element is drawn from a known distribution. 2 While numerous causality tests have been developed within this linearparametric framework, (see, for example,
The Size Problem of Bootstrap Tests when the Null isNon- or Semiparametric
Revista Colombiana de Estadistica
In non- and semiparametric testing, the wild bootstrap is a standard method for determining the critical values of tests. If the null hypothesis is also semi- or nonparametric, then we know that at least asymptotically oversmoothing is necessary in the pre-estimation of the null model for generating the bootstrap samples. See Hardle & Marron (1990, 1991). However, in practice this knowledge is of little help. In this note we highlight that this bandwidth choice problem can become quite serious. As an alternative, we briegly discuss the possibility of subsampling.