R help - Sep 2008 - Help with stl

I just realized after some tips and a little digging that what I was trying to
do "manually" has already been done. I was trying to fit my data using
'lm' then taking the "residual" data and trying to do a
spectral estimate (for seasonality) usiing fft and then passing the
"residual" of all of that to arima to get the irregular portion of the
time series forecast equation. I found 'stl' that advertises that it
does all of this (and probably better than my efforts). The only problem that I
found was that it takes a time-series object. Time-series objects don't
handle missing observations. Say I only have two observations for the year
(enough to fit a line through). I can easily fit a line through the observation
points but if I add in zeros for the missing observations least squares will
undoubtedly throw my observations out and fit the "wrong" line. The
other steps will more than likely have residual data so I don't have to
worry for these steps about missing data. Has anyone used 'stl' with
missing observations? If so what sort of tricks did you use to get around the
missing data and time series?

Thank you.

Kevin

if your two points are valid and you would be willing to run a least
spuares  regression on then you should be fairly comfortable with
using na.approx in the zoo package.  If interpolating values is going
to give you an erroneous result then maybe this is not the correct
path.  But have a look.

On Mon, Sep 1, 2008 at 1:50 AM,  <rkevinburton at charter.net>
wrote:
> I just realized after some tips and a little digging that what I was trying
to do "manually" has already been done. I was trying to fit my data
using 'lm' then taking the "residual" data and trying to do a
spectral estimate (for seasonality) usiing fft and then passing the
"residual" of all of that to arima to get the irregular portion of the
time series forecast equation. I found 'stl' that advertises that it
does all of this (and probably better than my efforts). The only problem that I
found was that it takes a time-series object. Time-series objects don't
handle missing observations. Say I only have two observations for the year
(enough to fit a line through). I can easily fit a line through the observation
points but if I add in zeros for the missing observations least squares will
undoubtedly throw my observations out and fit the "wrong" line. The
other steps will more than likely have residual data so I don't have to
worry for these steps about missing data. Has anyon!
>  e used 'stl' with missing observations? If so what sort of tricks
did you use to get around the missing data and time series?
>
> Thank you.
>
> Kevin
>
> ______________________________________________
> R-help at r-project.org mailing list
> 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.
>


-- 
Stephen Sefick
Research Scientist
Southeastern Natural Sciences Academy

Let's not spend our time and resources thinking about things that are
so little or so large that all they really do for us is puff us up and
make us feel like gods. We are mammals, and have not exhausted the
annoying little problems of being mammals.

	-K. Mullis

<rkevinburton <at> charter.net> writes:
> 
> I just realized after some tips and a little digging that what I was trying
to
do "manually" has already been
> done. I was trying to fit my data using 'lm' then taking the
"residual" data
and trying to do a spectral
> estimate (for seasonality) usiing fft and then passing the
"residual" of all
of that to arima to get the
> irregular portion of the time series forecast equation. I found
'stl' that
advertises that it does all of
> this (and probably better than my efforts). The only problem that I found
was
that it takes a time-series
> object. Time-series objects don't handle missing observations. Say I
only have
two observations for the
> year (enough to fit a line through). I can easily fit a line through the
observation points but if I add in
> zeros for the missing observations least squares will un
>  doubtedly throw my observations out and fit the "wrong" line.
The other steps
will more than likely have
> residual data so I don't have to worry for these steps about missing
data. Has
anyon!
>  e used 'stl' with missing observations? If so what sort of tricks
did you use
to get around the missing data
> and time series?
> 
> Thank you.
> 
> Kevin
> 
> ______________________________________________
> R-help <at> r-project.org mailing list
> 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.
> 
> 
Someone can correct me if I am wrong, but R's STL implementation 
uses the Fortran version of stl, which in interest of speed does not 
allow for missing observations.  If you use the stl implemented in S, 
you should be able to have missing observations.

If you read the original stl paper, the algorithm is described and is 
pretty straightforward to implement or do your own variation on.