R devel - Mar 2000 - A "stack" function

I am using R in my undergraduate engineering statistics courses, in
which I am currently discussing one-way analysis of variance.  Many of
the data sets for the exercises or examples give the data in the form
on one column per treatment level.  To use the aov or lm functions
such data should be converted to one column with all the response
observations and a companion column with indicators of the treatment
level.

In Minitab this operation is called "stacking" columns and the command
to perform the operation is "stack".  I have written a similar
function
for R.  Before taking the time to make it elegant and to document it,
I thought I would check if I have overlooked an existing way of doing
this.

Here is an example
> library(Devore5)
> data(ex10.09)
> boxplot(ex10.09)
> str(ex10.09)             # show the structure
`data.frame':	6 obs. of  4 variables:
 $ Wheat : num  5.2 4.5 6 6.1 6.7 5.8
 $ Barley: num  6.5 8 6.1 7.5 5.9 5.6
 $ Maize : num  5.8 4.7 6.4 4.9 6 5.2
 $ Oats  : num  8.3 6.1 7.8 7 5.5 7.2
> stack <- function(data)
+  data.frame(values = unlist(data),
+             lev = factor(rep(names(data), lapply(data, length))))
> stack(ex10.09)   # try it on the example
        values    lev
Wheat1     5.2  Wheat
Wheat2     4.5  Wheat
Wheat3     6.0  Wheat
Wheat4     6.1  Wheat
Wheat5     6.7  Wheat
Wheat6     5.8  Wheat
Barley1    6.5 Barley
Barley2    8.0 Barley
Barley3    6.1 Barley
Barley4    7.5 Barley
Barley5    5.9 Barley
Barley6    5.6 Barley
Maize1     5.8  Maize
Maize2     4.7  Maize
Maize3     6.4  Maize
Maize4     4.9  Maize
Maize5     6.0  Maize
Maize6     5.2  Maize
Oats1      8.3   Oats
Oats2      6.1   Oats
Oats3      7.8   Oats
Oats4      7.0   Oats
Oats5      5.5   Oats
Oats6      7.2   Oats
> fm1 <- lm(values ~ lev, stack(ex10.09))
> anova(fm1)
Analysis of Variance Table

Response: values
          Df  Sum Sq Mean Sq F value  Pr(>F)  
lev        3  8.9833  2.9944  3.9565 0.02293 *
Residuals 20 15.1367  0.7568                  
---
Signif. codes:  0  `***'  0.001  `**'  0.01  `*'  0.05  `.'  0.1
` '  1


-- 
Douglas Bates                            bates@stat.wisc.edu
Statistics Department                    608/262-2598
University of Wisconsin - Madison        http://www.stat.wisc.edu/~bates/
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-devel mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To:
r-devel-request@stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._

On 29 Mar 2000, Douglas Bates wrote:
> I am using R in my undergraduate engineering statistics courses, in
> which I am currently discussing one-way analysis of variance.  Many of
> the data sets for the exercises or examples give the data in the form
> on one column per treatment level.  To use the aov or lm functions
> such data should be converted to one column with all the response
> observations and a companion column with indicators of the treatment
> level.
> 
> In Minitab this operation is called "stacking" columns and the
command
> to perform the operation is "stack".  I have written a similar
function
> for R.  Before taking the time to make it elegant and to document it,
> I thought I would check if I have overlooked an existing way of doing
> this.
> 
It would be really nice if you could also do the reverse: take repeated
measurements in 'long' form and convert them into 'wide' form.
  In Stata both of these are available in the -reshape- command.  I don't
know any really simple way of doing it in R.

	-thomas

-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-devel mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To:
r-devel-request@stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._