Abstract: We propose a test to check heteroscedasticity and
investigate the limit behavior of the test statistic. The test can detect the
local alternative converging to the null at the parametric rate To calculate the critical values,
we employ the bootstrap resample technique. To investigate the numerical
performance, we conduct a simulation experiment to study the level and power of
the bootstrap test. We also compare the test with the test proposed by Dette and
Munk [J. Royal Statist. Soc.,
Series B 60 (1998), 693-708] in
the simulation study. The tests are further used to analyze one real data set.
The advantages of our test are, therefore, justified.
Keywords and phrases: bootstrap, Gaussian processes, heme oxygenase-1, carbon monoxide, local linear regression, semiparametric estimation.
