[R-sig-Finance] fSeries prob

[Spencer Graves]

Dear Øyvind Foshaug and Diethelm Wuertz:

ØYVIND:

	  Thanks for reporting this.  Could you please try the revised function 
below?  I changed three lines over the previous version

	  In case you are interested, I will outline here what I did to resolve 
this issue.  First, I looked on the 'armaFit' help page for an example. 
  Then I modified it in a way that successfully reproduced the error you 
reported:

xmpSeries("\nStart: Simulate an ARMA(2, 1) process > ")
x = armaSim(model = list(ar = c(0.5, -0.5), ma = 0.1), n = 1000)
Continue = xmpSeries("Press any key > ")
# Estimate the parameters:
fit = armaFit(x ~ arma(2, 1))
predict(fit, doplot=FALSE)

Error in predict.fARMA(fit, doplot = FALSE) :
	object "npred" not found

	  If you list 'predict', you will see it primarily consists of a call 
to 'UseMethod'.  The help page for 'UseMethod' says, "see also: 
methods".  When I tried 'methods("predict")', I got a character vector 
of length 23, all beginning "predict.".  This list included in 
particular, "predict.fARMA".  I listed that, made a local copy, found 
where 'npred' is designed and used, and made two minor changes in 
'predict.fARMA'.  After sourcing that, I got a different error, and I 
found an apparent patch for that.

	  Thanks again for reporting this problem.

DIETHELM:

	  Below please find a modification to your 'predict.fARMA' from fSeries 
221.10065.  I made three changes marked by comments.  Please let me know 
how you would like me to handle issues like this.  Until I hear from you 
I will send them to you like this -- unless I find a caret "^", which 
might be eaten by certain email software.

	  You might consider adding a 'predict(fit, doplot=FALSE)' to the help 
page example section.

	  Best Wishes,
	  Spencer Graves
##################################
sessionInfo()
Version 2.3.1 (2006-06-01)
i386-pc-mingw32

attached base packages:
[1] "methods"   "stats"     "graphics"  "grDevices" "utils"     "datasets"
[7] "base"

other attached packages:
     fSeries   fCalendar     fBasics
"221.10065" "221.10065" "221.10065"

	  Best Wishes,
####################################
predict.fARMA <-
function (object, n.ahead = 10, n.back = 50, conf = c(80, 95),
     doplot = TRUE, doprint = TRUE, ...)
{
     object = object at fit
     class(object) = object$class
     TS = "stats"
     if (object$tsmodel == "fracdiff") {
         warning("Prediction for FRACDIFF not yet implemented")
         return(NA)
     }
     options(warn = -1)
     pred = .predictTS(object = object, n.ahead = n.ahead, se.fit = TRUE)
     nint = length(conf)
     upper = lower = matrix(NA, ncol = nint, nrow = length(pred$pred))
     for (i in 1:nint) {
         qq = qnorm(0.5 * (1 + conf[i]/100))
         lower[, i] = pred$pred - qq * pred$se
         upper[, i] = pred$pred + qq * pred$se
     }
     colnames(lower) = colnames(upper) = paste(conf, "%", sep = "")
     shadecols = switch(1 + (length(conf) > 1), 7, length(conf):1)
     shadepalette = heat.colors(length(conf))
     col = 1
     data = as.ts(object$x)
     freq = frequency(data)
     start = start(data)
     n = length(data)
     pred.mean = pred$pred
     npred = length(pred.mean)
#   The preceeding 2 lines (pred.mean and npred)
#   were moved from after if(doplot) to before.
     if (doplot) {
         ylim = range(c(data[(n - n.back + 1):n], pred.mean),
             na.rm = TRUE)
         ylim = range(ylim, lower, upper, na.rm = TRUE)
         ylab = paste("Series: ", object$series)
         plot(ts(c(data[(n - n.back + 1):n], pred.mean[1], rep(NA,
             npred - 1)), end = tsp(data)[2] + npred/freq, f = freq),
             ylim = ylim, ylab = ylab)
         title(main = paste(object$tstitle))
     }
     xx = tsp(data)[2] + (1:npred)/freq
     idx = rev(order(conf))
     if (nint > 1)
         palette(shadepalette)
#   "if(doplot)" was added before 'for(i in 1:nint)'
     if(doplot)for (i in 1:nint) {
         polygon(c(xx, rev(xx)), c(lower[, idx[i]], rev(upper[,
             idx[i]])), col = shadecols[i], border = FALSE)
     }
     palette("default")
     lines(ts(pred.mean, start = tsp(data)[2] + 1/freq, f = freq),
         lty = 1, col = 4)
     nconf = length(conf)
     out = pred.mean
     upper = as.matrix(upper)
     lower = as.matrix(lower)
     names = "Forecast"
     for (i in nconf:1) {
         out = cbind(out, lower[, i])
         names = c(names, paste("Low", conf[i]))
     }
     out = cbind(out, pred.mean)
     names = c(names, "Forecast")
     for (i in 1:nconf) {
         out = cbind(out, upper[, i])
         names = c(names, paste("High", conf[i]))
     }
     out = round(out, digits = 4)[, 2:(2 * nconf + 2)]
     colnames(out) = names[2:(2 * nconf + 2)]
     if (doprint)
         print(out)
     options(warn = 0)
     result = list(pred = pred$pred, se = pred$se)
     invisible(result)
}
####################################
Øyvind Foshaug wrote:
> Hi,
>   I seem to get an error when I apply predict with the 
option doplot=FALSE in function armaFit. It says npred not found.
>    
>   Thanks,
>   Oyvind Foshaug
> 
>  		
> ---------------------------------
> 
> 
> 	[[alternative HTML version deleted]]
> 
> _______________________________________________
> R-SIG-Finance at stat.math.ethz.ch mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-finance