Newly developed and advanced methods for nonlinear time series analysis are in general not available in standard software packages. Moreover, their implementation requires substantial time, computing power as well as programming skills. The recent results on lag and bandwidth selection methods for nonlinear autoregressive time series models provide such a scenario. Application of these methods requires enormous computing resources if larger samples are considered. In this paper we suggest a method to provide empirical researchers with a fast access to new methods as well as to powerful computing environments. It is illustrated with a recently suggested nonparametric lag selection procedure based on CAFPE (Corrected Asymptotic Final Prediction Error). Our approach is based on the XploRe quantlet technology. Its worldwide Web usage is made possible by a specific client/server architecture. It allows researchers to use the quantlet computing service without knowing either the statistical computing language or the server location. Quantlets are accessed via standard WWW browsers or via a Java client which works like a standard desktop environment. This architecture allows a flexible scaling of time consuming computations on either client or server. The XploRe quantlet service is helpful in constructing research books and interactive teaching environments as the electronic version of this paper demonstrates.
distributed computing lag selection nonparametric time series analysis quantlets web based computing