On Sun, 25 Apr 2004 kjetil at acelerate.com wrote:
> Is this a bug?
Without even a traceback() nor reproducible code it is hard to say.
But for lm, you don't want time series and it would be better to drop the
ts class and tsp attributes.
> seasonal.dummies <-
> function(x, contr=NULL) {
> # takes a time series and returns a matrix of seasonal dummies for
> # x. This is almost cycle(x), we only have to make it into a factor
> # and add suitable level names.
> # return a matrix which includes a constant!
> # level names here assumes frequency is 12!
> cyc <- factor( cycle(x), labels=c("ene","feb",
"mar", "abr", "may",
> "jun","jul", "ago",
"sep", "oct","nov", "dic") )
> if(!is.null(contr)) cyc <- C(cyc, contr)
> model.matrix( ~ cyc )
> }
>
>
> > modelo1 <- function(xx, yy) {
> # xx es serie explicativa (economia)
> # yy es serie para predecir --- gasolina.eq
> name.x <- deparse(substitute(xx))
> name.y <- deparse(substitute(yy))
> xx <- ts(rep(xx,rep(3, length(xx))), frequency=12, start>
c(start(xx)[1], (start(xx)[2]-1)*3+1) )
> common <- ts.intersect(xx, yy)
> colnames(common) <- c(name.x, name.y)
> xx <- common[,1]
> yy <- common[,2]
> sea <- seasonal.dummies(xx)
> return(lm(yy ~ 0+sea+xx))
> }
> > mod1 <- modelo1(PIBtrimestral, gasolina.eq)
> Error in "storage.mode<-"(`*tmp*`, value = "double")
:
> invalid time series parameters specified
>
> lm() works with time series objects made directly with ts(), and the
> arguments to modelo1() is of this type.
>
> Kjetil Halvorsen
>
> ______________________________________________
> R-help at stat.math.ethz.ch mailing list
> https://www.stat.math.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide!
http://www.R-project.org/posting-guide.html
>
>
--
Brian D. Ripley, ripley at stats.ox.ac.uk
Professor of Applied Statistics, http://www.stats.ox.ac.uk/~ripley/
University of Oxford, Tel: +44 1865 272861 (self)
1 South Parks Road, +44 1865 272866 (PA)
Oxford OX1 3TG, UK Fax: +44 1865 272595