[R] complete newbie Q

[Jonathan Baron]

On 06/14/04 09:25, jeroen clarysse wrote:
>Hi all
>
>I'm a programmer at the psychology dept, and last week, I was asked to write
>an application to analyze some result data from CO2 measurement experiments.
>I don't want to reinvent the wheel, so before I start custom coding in C,
>I'd though to look around a bit and bumped into R on freshmeat.
>
>basically, it is a 2-column data sheet, with timings on col1 and CO2 value
>on col2. These value have a pretty nice oscillating nature, with some
>occasional false spikes.
>
>the analysis simply means extracting the 'ceilings' of the curves = the
>start-end times of the top of each oscilation.
>
>my Q to the mailinglist is now : can such analysis be done in R ? Or is R
>not the appropriate package for this kind of stuff ?

R is certainly the appropriate package - especially for someone
with programming experience - but this is not a simple problem no
matter what you use.  The problem is to eliminate the noise, the
"false spikes."  I cannot give you the solution, and I notice
that nobody else has replied either.  But I can tell you that I
dealt with a very similar problem (helping the Mozilla Foundation
measure the speed of page loading, which turned out to have a
periodicity) using various time-series functions, such as acf, as
well as trimmed means to get rid of outliers.  But the situation
was a bit different, as there were several observations at each
time point, so I could apply the trimmed mean to that time point.
Unfortunately, I don't have time to get more involved in your
problem, but this may get you started.  I think what you might
have to do is iterate between fitting the model and eliminating
outliers from the residuals, but maybe some statistician on this
list will have a better idea.

Jon
-- 
Jonathan Baron, Professor of Psychology, University of Pennsylvania
Home page:            http://www.sas.upenn.edu/~baron