R help - Dec 2004 - graphics

Dear R-user,
I am a R beginner, and therefore my questions are very basic.
I have a simple problem: I would like to plot 100 time series each 
containing 55 steps. The data are stored in a matrix of 100 columns and 
55 rows. The first problem is to load the data from a file: I tried the 
read.table(), the scan() and the matrix(scan()) options, but I have 
problems to allocate the single columns. The list() option could be a 
solution, but it is very unconvenient: list(0,0,0......).
And how do I plot a single time series, let's say the 50s? And how to 
plot all of them?

The last problem is maybe more advanced: I would like to plot all 100 
time series, but with a confidence interval, where the density of data 
is indicated by the density of the colour of the confidence interval. A 
colleague gave me the tip that this is possible in R2.0 and described in 
the newest newsletter ...

Many thanks in advanced,
Fred

Fred Hattermann wrote:
> 
> Dear R-user,
> I am a R beginner, and therefore my questions are very basic.
> I have a simple problem: I would like to plot 100 time series each 
> containing 55 steps. The data are stored in a matrix of 100 columns and 
> 55 rows. The first problem is to load the data from a file: I tried the 
> read.table(), the scan() and the matrix(scan()) options, but I have 
> problems to allocate the single columns. The list() option could be a 
> solution, but it is very unconvenient: list(0,0,0......).
> And how do I plot a single time series, let's say the 50s? And how to 
> plot all of them?

These questions are really basic, they are covered in the manuals and 
any good book about R. Please read at least
a) the manual "An Introduction to R"
b) the manual "R Data Import/Export"
c) the posting guide (see below, has been appended at the end of your 
message)

After you have read through and tried to solve your problem again, you 
might want to come up with a specific question again.

> The last problem is maybe more advanced: I would like to plot all 100 
> time series, but with a confidence interval, where the density of data 
> is indicated by the density of the colour of the confidence interval. A 
> colleague gave me the tip that this is possible in R2.0 and described in 
> the newest newsletter ...
This question is also unspecific. Which kind of confidence interval? 
What does "density of data" mean here? Do you mean a global quantity
for
the whole series or really the empirical (estimated) density so that 
color changes from point to point?

Uwe Ligges

> Many thanks in advanced,
> Fred
> 
> ______________________________________________
> R-help at stat.math.ethz.ch mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide! 
> http://www.R-project.org/posting-guide.html

Dear Jim,
many thanks for your reply and support.
It seems to be that with your help I could solve my problem with the 
plotting of the data. The only thing that does not work is to see the 
coloured lines, maybe because of the crowd of curves.
With the density of curves I ment to distinguish in the confidence 
interval between areas, where a lot of curves (lines) are located, and 
areas having a lower density of lines. In other words, I would like to 
show the area, where 100 % of the lines are included, then the area, 
where 90 % are included, and so forth with the other quantiles, but with 
a steady change of colours, and not by a stepwise change (if posible).

Many thanks again,
Fred


Jim Lemon wrote:
>Fred Hattermann wrote:
>  
>
>>Dear R-user,
>>I am a R beginner, and therefore my questions are very basic.
>>I have a simple problem: I would like to plot 100 time series each
>>containing 55 steps. The data are stored in a matrix of 100 columns and
>>55 rows. The first problem is to load the data from a file: I tried the
>>read.table(), the scan() and the matrix(scan()) options, but I have
>>problems to allocate the single columns. The list() option could be a
>>solution, but it is very unconvenient: list(0,0,0......).
>>    
>>
>
># generate some random numbers
>testts.df<-data.frame(matrix(rnorm(5500)/5,nrow=55))
># superimpose them on a sine curve
>newts<-sapply(testts.df,function(x) return(x+sin(seq(0,pi*2,length=55))))
># make it a time series
>newts<-as.ts(newts)
># write out the data
>write.table(newts,"newts.dat")
># read it in again
>newts<-as.ts(read.table("newts.dat"))
>
>  
>
>>And how do I plot a single time series, let's say the 50s? And how
to
>>plot all of them?
>>
>>    
>>
># plot the first one
>plot(newts[,1],ylim=range(newts))
># add the other 99 lines - probably pretty messy!
>for(i in 2:100) lines(newts[,i])
>
>  
>
>>The last problem is maybe more advanced: I would like to plot all 100
>>time series, but with a confidence interval, where the density of data
>>is indicated by the density of the colour of the confidence interval.
>>    
>>
>
># get the means of the observation points
>newts.means<-apply(as.matrix(newts),1,mean)
># calculate a CI - probably not the one you want
>newts.ci<-1.96*sapply(as.matrix(newts),sd)
># plot the CI
>lines(newts.means+newts.ci,col="red")
>lines(newts.means-newts.ci,col="green")
>
>I'm not sure what you mean by the "density" of the curve, so I
can't suggest
>anything. However, I am adding a function named "color.scale" to
the next
>version of the "plotrix" package, so I'll email you when I put
it up on CRAN.
>
>Jim
>
>
>  
>

-- 
 *******************************************************
 * Potsdam Institute for Climate Impact Research (PIK) *
 * Telegrafenberg  C 4   D-14473 Potsdam               *
 * PO box:  60 12 03     D-14412 Potsdam      	       *
 * Tel.:  (0331) 288 - 2649                            *
 * Fax:   (0331) 288 - 2695                            *
 * e-mail: hattermann at PIK-Potsdam.de                   *
 *******************************************************