S-PLUS-functions for likelihood ratio tests in the linear mixed model
- This software is no longer supported. Of course, you are still welcome to use it and to send me questions about it, but the answers might not be very informative and I won`t spend any time on changing the code / functionality.
- Description:
Representing generalised geoadditive mixed models as generalised linear mixed models does not only allow to estimate smoothing parameters via (restricted) maximum likelihood, but also to perform likelihood ratio tests on this smoothing parameters if the model is not too complicated and if the response is normally distributed. As special cases one obtains tests of linearity in nonparametric regression models or tests for the presence of spatial effects. Since under H0 the smoothing parameter is on the boundary of the parameter space and the observations may no longer be viewed as independent (at least under the alternative), the standard theory for likelihood ratio tests is not adequate here. A more general theory has been developed by Crainiceanu and Ruppert (2002) to deal with this problem. To get a better understanding of their results I developed some S-PLUS-functions that allow to compute certain quanitites of the distribution of the likelihood ratio test statistics. Moreover the asymptotic distribution of these test statistics may be simulated.
- Download:
