[R] HoltWinters function

[Nichenametla, Karthik K]

Hi David,

I am facing some issues with the HoltWinters function in R (v2.9.2). I am doing time series analysis using this method. For the time series data that I used, (alpha,beta,gamma) parameters came out to be (1,0,0). I think this theoretically should mean that the predicted timeseries values will be constant. But the predicted values were not coming out to be constant. What does this mean ?

One more point is that when I ran the same code in older version(v2.8.x), I got different parameter values and different predicted values. Was there any change in HoltWinters function?

Below is my R window text:
########################################
> a=as.matrix(read.table("embedded_mani1.txt"))
>
> K=dim(a)[2]
> N=dim(a)[1]
>
> ts1=ts(a, frequency=7, start=c(1,1))
> #### Holtwinters with start.period=7
> params = matrix(nrow=3,ncol=K)
> prediction=matrix(0,nrow=K,ncol=28)
> ts_result=HoltWinters(ts1[,1], start.periods =7)
> params[,1]=c(ts_result$alpha,ts_result$beta,ts_result$gamma)
> print(params)
     [,1] [,2] [,3] [,4]
[1,]    1   NA   NA   NA
[2,]    0   NA   NA   NA
[3,]    0   NA   NA   NA
> pre=predict(ts_result,28,se.fit=FALSE)
> head(pre)
[1] 0.2703228 1.0252342 1.4506853 1.3881035 1.3030489 1.0624612
####################################################

Please give your comments on this.

Regards,
Karthik,
Graduate Student,
Georgia Tech.