similar to: corrupted smoothing kernel ?

Hi, I am far from experienced in both R and time series hence the question. The code for spec.pgram() seems to involve a circularity of the kernel (see below) yielding new power estimates to all frequencies computed by FFT. " if (!is.null(kernel)) { for (i in 1:ncol(x)) for (j in 1:ncol(x)) pgram[, i, j] <- kernapply(pgram[, i, j], kernel, circular = TRUE)

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

Code to reproduce > kernel("dirichlet") Error in kernel("dirichlet") : argument "r" is missing, with no default The help says r??? the kernel order for a Fejer kernel. Thanks to update the help. Best, Samuel

Hi all, as I am trying to move slowly from just "working" to "good" code, I'd like to ask if there's a smarter way than using a for-loop in tasks like the example below. I need to obtain the extrema of the cumulated sum of a detrended time series. The following code is currently used, please have a look at the comments for my questions and remarks: system.time({ X

Hi All, What do you consider to be the best book(reference) on nonparametric regression? I am currently reading the book of Kunio Takezawa(2006): "Introduction to nonparametric regression". Is the book of Hardle(1990): "Applied nonparametric regression" better? or maybe another book? This is off topic, but most of the books is using R or S-plus. Thanks Hennie

I need some real help on this, really stuck how are the coefficients for ur.ers(y, type = c("DF-GLS", "P-test"), model = c("constant", "trend"), lag.max = 0) The max lag is set at zero, so the regression should simply be Diff(zt) = a*z(t-1) where a is the value i'm trying to find and z(t)'s are the detrended values. but through performing

Hello - I am currently simulating bivariate AR(1) time series data and have the following line in my code: Px=spec.pgram(ts.union(X,XX),spans=c(?,?)) The spans option is where I enter in the vector containing the Daniell smoother numbers, but I don't know what a Daniell smoother is (hence the question marks). Can somebody please tell me? Is there another option where I can simply enter in

Dear all, This is potentially a bug in spec.pgram, when the number of samples is odd,spec.pgramreturns a different result withfast = TRUE, the example below contains the two varieties with a reference spectrum calculated manually. the number of returned spectra is also larger (50 compared to 49) whenfast = TRUE x <- rnorm( 99 ) plot(spec.pgram(x, taper = 0 , detrend = FALSE , plot =

Dear R users I have two questions about estimating the spectral power of a time series: 1) I came across a funny thing with the following code: data(co2) par(mfrow=c(2,1)) co2.sp1<-spectrum(co2,detrend=T,demean=T,span=3) co2.sp2<-spectrum(co2[1:468],detrend=T,demean=T,span=3) The first plot displays the frequencies ranging from 0 to 6 whearas the second plot displays the same curve but

Hello! I wonder if there is a typo in detrending code of spec.pgram in spectrum.R from stats package. One can see in the code https://svn.r-project.org/R/trunk/src/library/stats/R/spectrum.R . I am afraid there is a typo and the code should look like if (detrend) { t <- 1L:N - (N + 1)/2 sumt2 <- N * (N^2 - 1)/12 for (i in 1L:ncol(x)) x[, i] <- x[, i] -

Hi All, I am a PhD student in forestry science and I am working with LiDAR data set (huge data set). I am a brand-new in R and geostatistic (SORRY, my background it?s in forestry) but I wish improve my skill for improve myself. I wish to develop a methodology to processing a large data-set of points (typical in LiDAR) but there is a problem with memory. I had created a subsample data-base but

Hello, I propose two small changes to spec.pgram to get modest speedup when dealing with input (x) having multiple columns. With plot = FALSE, I commonly see ~10-20% speedup, for a two column input matrix and the speedup increases for more columns with a maximum close to 45%. In the function as it currently exists, only the upper right triangle of pgram is necessary and pgram is not returned by

Hi, Are there any functions in R that could be used to estimate the phase-shift between two semi-sinusoidal vectors? Here is what I have tried so far, using the spectrum() function -- possibly incorrectly: # generate some fake data, normalized to unit circle x <- jitter(seq(-2*pi, 2*pi, by=0.1), amount=pi/8) # functions defining two out-of-phase phenomena f1 <- function(x)

I have a time series of 1095 data corresponding to a daily data of three years. I want to know how to use ma(timeserie, order=??, centre=??) to detect the trend: which order is suitable and what is the difference between centre= true or false. How to avoid these errors: 1-Error in timeserie - trend : ? argument non num?rique pour un op?rateur binaire="non-numeric argument for a binary

I am reading some documentation about Cross Spectrum Analysis as a technique to compare spectra. My understanding is that it estimates the correlation strength between quasi-periodic structures embedded in two signals. I believe it may be useful for my signals analysis. I was referred to the R functions that implement this type of analysis. I tried all the examples which generated a series of

Hi, Hope this does not sound too ignorant . I am trying to detrend and transform variables to achieve normality and stationarity (for time series use, namely spectral analysis). I am using the boxcox transformations. As my dataset contains zeros, I found I need to add a constant to it in order to run "boxcox". I have ran tests adding several types of constants, from .0001

I have a question regarding the adf.test command in the tseries library. I have a vector of time series observations (2265 daily log prices for the OEX to be exact). I also have this same data in first-differenced form. I want to test both vectors individually for staionarity with an Augmented Dickey-Fuller test. I noticed when I use the adf.test command from the tseries library, the general

There is now a preliminary version of a time series package in the R-devel snapshots, and we would welcome feedback on it. It is based in part on the packages bats (Martyn Plummer) and tseries (Adrian Trapletti) and in part on code I had or have written. (Thanks for the contributions, Martyn and Adrian!) Some of the existing ts code has been changed, for example to plot multiple time series, so

I'm doing convolutions in the frequency domain and need to know the layout of the Fourier modes returned by fft. (This is leading up to a more involved question about moment generating functions, but I need to know if I've got this part correct first.) I think in 1D the pattern is: 0 1 2 3 -2 1 (even) 0 1 2 3 -3 2 1 (odd) In 2D is it simply (for a square matrix): 0 1 2 -1 (horizontal)

hello, I have been using the per function in package longmemo to obtain a simple raw periodogram. I am considering to switch to the function spec.pgram since I want to be able to do tapering. To compare both I used spec.pgram with the options as suggested in the documentation of per {longmemo} to make them correspond. Now I have found on a variety of examples that there is a shift between