[R] Understanding TS objects

[Achim Zeileis]

On Tue, 13 Mar 2018, JEFFERY REICHMAN wrote:

> R Help Community
>
> I'm trying to understand time series (TS) objects.  Thought I understood 
> but recently have run into a series of error messages that I'm not sure 
> how to handle.  I have 15 years of quarterly data and I typically create 
> a TS object via something like...
>
> data.ts <- ts(mydata, start = 2002, frequency = 4)
>
> this create a matric as opposed to a vector object

This depends on what "mydata" is which you haven't shown...

If "mydata" is a univariate vector, everything works ok:

mydata <- rnorm(15 * 4)
data.ts <- ts(mydata, start = 2002, frequency = 4)
data.stl <- stl(data.ts, "periodic")

However, if "mydata" is a matrix, e.g., a 3-column matrix in the example 
below:

mydata <- matrix(rnorm(15 * 4 * 3), ncol = 3)

then the error occurs.

Furthermore, the same problem will occur if mydata is 1-column matrix or a 
1-column data frame.

> as I receive a univariate error when I try to decompose the data using 
> the STL function
>
> data.stl <- stl(data.ts, "periodic")
> Error in stl(data.ts, "periodic") : only univariate series are allowed
>
> ok so
>
> is.vector(data.ts)
> [1] FALSE

This is always FALSE for a "ts" object, even if it is univariate, because 
it has further attributes, namely the time-series properties (tsp).

> so to convert to a vector I'll use
> data.ts <- as.vector(data.ts)

This will drop the "ts" class.

The cleanest way is probably to create a vector "mydata" and then the 
univariate "data.ts".

hth,
Z

> but then I lose the frequency as the periods as the data becomes frequency = 1
> data.ts <- stl <- stl(data.ts, "periodic")
> Error in stl(data.ts, "periodic") :
>   series is not periodic or has less than two periods.
>
> So am I missing a  parameter or is there a more general/proper way to create a time series object? First time I've run into this problem .  I can always decompose  via an alternative methods so there are work arounds.  But just trying to understand what I'm not doing programmatically at this point.
>
> Jeff Reichman
>
> ______________________________________________
> R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
>