arima.mle() will calculate the variance-covariance matrix for the autoregressive and moving average coefficients (citing the help-page). Regards, Henrik -- Henrik Aalborg Nielsen Informatics and Mat
Hi Kristian There is also a solution which does not rely on as.numeric returning the result in days: (tDdata$Event-tDdata$DOB) / timeSpan("365 days 6 hours") # or "365d 6h" Regards, Henrik -- Henrik
Models like y(t) = a + b*x(t) + AR(2)-error : arima.mle (xreg=cbind(1,x)) Models like y(t) = a0 + a1*y(t-1) + a2*y(t-2) + b*x(t) + i.i.d. gaussian error : Set up the variables in a data frame and fit
Hi John, Your problem is caused by the scooping rules. The following will work and .tmp.dat will not exist when loess.fitter returns. I think 'get(...)' can just be replaced by '.tmp.dat', but 'get'
I believe do.call("+", matrices) will do the job. /Henrik -- Henrik Aalborg Nielsen Informatics and Mathematical Modelling Section for Statistics, Time Series Group Technical University of Denmark Ri
Dear Barry, The time must be a numeric, e.g. time <- timeSeq(from = "1/1/2003", to = "1/31/2003", by = timeSpan("6h")) y <- rnorm(n = length(time)) loess(y ~ time) # Don't work daysFromStart <- (time
Dear John and the rest of S-news, Sorry - I did not check my suggestion enough - only with a list of length two ... /Henrik -- Henrik Aalborg Nielsen Informatics and Mathematical Modelling Section fo
Dear Bevan, As you noticed it is not possible to bootstrap something which is not i.i.d. I would look for an appropriate ARMA-model and then simulate new series using this model (functions arima.mle,
One posibility is to transform the columns into time spans, e.g.: timeSpan(c("10:30", "09:01"), in.format="%H:%M") and then just subtract the one from the other. The output is still a time span; usin
The problem seems to be that the stuff is there but it is very hard to find. When I needed a function to concat a number of timeSeries I searched the help and timeSeries section of the manual, but th
Dear Kelly, I think that the most simple solution is to write a function which takes arguments time and Delta; say ff (more independent variables?). The call to nls would then be: nls(vl ~ ff(Delta,
Dear all, On the help page for arima.fracdiff it is stated thet the loglik element returned by arima.fracdiff is "a quantity proportional to the logarithm of the maximum likelihood (up to an additive
Dear Paul, Maybe the following function is helpful - it concats timeSeries or signalSeries directly: seriesConcat <- function(..., list.series = list(...)) { elemClass <- unlist(lapply(list.series, c
Could someone from Insightful please comment on the current state and plans for the Matrix-library; it is in the 6.0 and 6.1 Solaris and Linux versions, but from Steve's email it seems that it is not
Dear S-list, I use timeDate, timeSpan, and timeSeries at a regular basis for handling (highly) irregular timeSeries and I do think that IT WORKS VERY WELL, the exceptions being that plotting highly i