Gorgens, Jakob (Tue)2015-12-130167-7152http://hdl.handle.net/1885/75055This paper proposes a new nonparametric test for the hypothesis that the regression functions in two or more populations are the same. The test is based on local linear estimates using data-driven bandwidth selectors. The test is applicable to data with random regressors and heteroskedastic responses. Simulations indicate the test has good power.Keywords: Bias correction; Local polynomial fitting; Nonparametric testing200210.1016/S0167-7152(02)00283-32015-12-11