R help - Feb 2008 - Help with package reshape, wide to long

Hello,

I am having difficulty figuring out how to use functions in the 
reshape package to perform a wide to long transformation

I have a "wide" dataframe whose columns are like this example:

   id1 id2 subject treat height weight age

id1 and id2 are unique for each row
subject and treat are not unique for each row
height, weight, and age are different types of measurements made on 
each unique combination of subject and treatment

I want to reshape to a long format which will look like this:

  id1 id2 subject treat measurement.type value

where

measurement.type identifies the type of measurement, i.e. 'height', 
'weight', 'age'
value contains the values of those measurements
and the other variables are replicated as necessary

To put it another way, can I use reshape() to transform my original 
dataframe, which has 45 rows, into a "long" form that has 3*45  = 135 
rows:  45 rows for height, 45 for weight, 45 for age, with the other 
variables carried along as is within each set of 45, and a new 
variable that identifies the type of measurement in each row of the 
long form, i.e., 'height' in 45 rows, 'weight' in 45 rows, and
'age'
in 45 rows.

I know it's not difficult to do this with explicit looping, using 
rbind(), but it seems like reshape() is a natural tool. But I'm not 
getting it, so I'd appreciate some help.

(in case anyone is wondering whether it makes sense to do this, 
height, weight, and age are just examples; it makes more sense with 
my actual measurements)

Thanks
-Don
-- 
--------------------------------------
Don MacQueen
Environmental Protection Department
Lawrence Livermore National Laboratory
Livermore, CA, USA
925-423-1062

On 2/7/2008 11:40 AM, Don MacQueen wrote:
> Hello,
> 
> I am having difficulty figuring out how to use functions in the 
> reshape package to perform a wide to long transformation
> 
> I have a "wide" dataframe whose columns are like this example:
> 
>    id1 id2 subject treat height weight age
> 
> id1 and id2 are unique for each row
> subject and treat are not unique for each row
> height, weight, and age are different types of measurements made on 
> each unique combination of subject and treatment
> 
> I want to reshape to a long format which will look like this:
> 
>   id1 id2 subject treat measurement.type value
> 
> where
> 
> measurement.type identifies the type of measurement, i.e. 'height',
> 'weight', 'age'
> value contains the values of those measurements
> and the other variables are replicated as necessary
   Does something like this work for you?

df <- data.frame(id1 = c(1,2,3),
                  id2 = c(4,5,6),
                  subject = runif(3),
                  treat = runif(3),
                  height = runif(3),
                  weight = runif(3),
                  age = runif(3))

library(reshape)

melt(df, measure.var = c("height","weight","age"))

   id1 id2   subject      treat variable      value
1   1   4 0.4736760 0.02221288   height 0.09290354
2   2   5 0.1708840 0.84465476   height 0.70660286
3   3   6 0.6955085 0.56084004   height 0.52761841
4   1   4 0.4736760 0.02221288   weight 0.08853412
5   2   5 0.1708840 0.84465476   weight 0.85745906
6   3   6 0.6955085 0.56084004   weight 0.29902097
7   1   4 0.4736760 0.02221288      age 0.91842595
8   2   5 0.1708840 0.84465476      age 0.93592277
9   3   6 0.6955085 0.56084004      age 0.01587108
> To put it another way, can I use reshape() to transform my original 
> dataframe, which has 45 rows, into a "long" form that has 3*45  =
135
> rows:  45 rows for height, 45 for weight, 45 for age, with the other 
> variables carried along as is within each set of 45, and a new 
> variable that identifies the type of measurement in each row of the 
> long form, i.e., 'height' in 45 rows, 'weight' in 45 rows,
and 'age'
> in 45 rows.
> 
> I know it's not difficult to do this with explicit looping, using 
> rbind(), but it seems like reshape() is a natural tool. But I'm not 
> getting it, so I'd appreciate some help.
> 
> (in case anyone is wondering whether it makes sense to do this, 
> height, weight, and age are just examples; it makes more sense with 
> my actual measurements)
> 
> Thanks
> -Don
-- 
Chuck Cleland, Ph.D.
NDRI, Inc.
71 West 23rd Street, 8th floor
New York, NY 10010
tel: (212) 845-4495 (Tu, Th)
tel: (732) 512-0171 (M, W, F)
fax: (917) 438-0894