[R] RES: another aov results interpretation question

[Gabor Grothendieck]

On 6/20/05, Paulo Brando <pmbrando at ipam.org.br> wrote:
> 
> Dear All,
> 
> I created a script to calculate averages - two groups: "parcel" and
> "date" - and, based on these averages, make a graph. The problem is that
> 'R' does not recognize the first column even if I try to insert one.
> 
> A brief example
> 
> Raw data:
> 
> Data <- sample(1:100, 30, replace = FALSE, prob = NULL)
> Date <- rep(c("02/30/2004","03/15/2004", "04/16/2004", "06/14/2004",
> "07/08/2004"), 6)
> Parcel <- rep(c("DRY", "CONTROL"),15)
> 
> Df <- data.frame(Parcel, Date, Data)
> 
> Res <- tapply(Df$Data, list(Date = Df$Date, Parcel = Df$Parcel), mean)
> 
> 
> > Res
>            Parcel
> Date          CONTROL      DRY
>  02/30/2004 53.00000 52.33333
>  03/15/2004 54.00000 67.66667
>  04/16/2004 54.66667 30.00000
>  06/14/2004 27.66667 20.00000
>  07/08/2004 59.00000 38.00000
> 
> > colnames(Res)
> [1] "CONTROL" "DRY"
> 
> > matplot(Res[,1], Res[,-1], type = "l")
> 
> ### It does not recognize the colunm "Date". Why?

Date is a not a column in the matrix -- it is the name
of the row dimension (and Parcel is the name of the
column dimension).   Res is a 5 by 2 matrix whose
colnames are CONTROL and DRY.  Try this:

   dimnames(Res)
   dim(Res)  # 5 by 2

Also note that the first date (Feb 30th) is illegal and matplot
does not work with dates in any case.  Fix up the erroneous
date and convert the row names to an object, xx, of chron or Date 
class, using plot for the axes and matlines for the lines:

   xx <- format(rownames(Res), "%m/%d/%Y")  # Date class
   plot(range(xx), range(Res), type = "n")  # axes
   matlines(xx, Res)

More on dates is in R News 4/1.