mauede at alice.it
2009-May-17 16:38 UTC
[R] Some questions about package "pastecs" and "stats"
My goal is to remove signals trend without any a-priori knowledge of the trend type, if any. Some signals are very noisy and non-stationary (example attached). I have experimented with a number of techniques. Staring at the results, I can hardly tell which method is best. I am attaching the result of function "local.trend" as I cannot understand it. Nor I can make a sense of the returned numerical data. I also tried function "trend.test". Neither one help me detect the presence of non-linear trend. I would appreciate some suggestions / answers / literature refereces (possibly in English) about the following problems (I am in the process of learning): 1. General question: What are the tests that reveal the presence of trend, besides Kendall-Tau, ACF, Lagged Scatterplot ? 2. General question: Is there any way to rank the goodness of trend removal ? In short, how can I know I have removed all the trend (the residual is a trend-free signal)? For instance, a package implementing SSA, sometimes enters an endless loop. It keeps detecting trend in the residual signal no matter how many times the removing procedure is run and the signal reconstructed. 3. Question about package "pastecs": Are there any criteria helping with the choice of proper values for parameters "order", "times" , "type" for functions "decaverage" and "decmedian" ? 4. Question about package "stats": Are there any criteria helping with the choice of proper kernel shape ("daniell", "dirichlet", "fejer", "modified.daniell") and order and dimension ? Thank you in advance. Maura e tutti i telefonini TIM! Vai su -------------- next part -------------- A non-text attachment was scrubbed... Name: 2425-LocalTrend.pdf Type: application/pdf Size: 238370 bytes Desc: 2425-LocalTrend.pdf URL: <https://stat.ethz.ch/pipermail/r-help/attachments/20090517/06e3e221/attachment-0002.pdf>