Evaluating a Series-Based Semiparametric Test for Additive Separability
University of Copenhagen - Department of Economics
Jesper Riis-Vestergaard Sorensen
University of California, Los Angeles (UCLA) - Department of Economics
Young, Conaway, Stargatt & Taylor
July 20, 2011
Empiria Working Paper No. 5
In this paper we perform a sensitivity analysis of a test for additive separability proposed by through a number of Monte Carlo studies. We evaluate the relative performance of using series base functions with a global focus, namely polynomials, compared to using the linear spline base employed in the original article, which has a local focus. Focusing on the size of the test, we find that the two bases perform equally well as long as the series length (the number of approximating functions) is chosen appropriately. Generally, the test seems more robust to the issue of oversmoothing than, for example, a test considered in Wooldridge (1992). Furthermore, we try various data-generating processes and find the test to be robust to non-uniformly distributed explanatory variables and discontinuity in the unknown function. We conclude that the test seems highly robust to many of the issues appearing in economic applications.
Number of Pages in PDF File: 11
JEL Classification: C12, C14working papers series
Date posted: November 20, 2011
© 2013 Social Science Electronic Publishing, Inc. All Rights Reserved.
This page was processed by apollo5 in 0.329 seconds