Data.frame manipulation

Hi All,

I'm conducting a meta-analysis and have taken a data.frame with 

rows per
study (for each effect size) and performed a weighted average of 

size for
each study. This results in a reduced # of rows. I am particularly
interested in
simply reducing the additional variables in the data.frame to the 

of the
corresponding id variable. For example:

id<-c(1,2,2,3,3,3)
es<-c(.3,.1,.3,.1,.2,.3)
mod1<-c(2,4,4,1,1,1)
mod2<-c("wai","other","calpas","wai","itas","other")
data<-as.data.frame(cbind(id,es,mod1,mod2))

data

   id   es    mod1 mod2
1  1   0.3    2     wai
2  2   0.1    4     other
3  2   0.2    4     calpas
4  3   0.1    1     itas
5  3   0.2    1     wai
6  3   0.3    1     wai

# I would like to reduce the entire data.frame like this:

id  es   mod1  mod2

1  .30     2        wai
2  .15     4        other
3  .20     1         itas

# If possible, I would also like the option of this (collapsing on id 

mod2):

id  es   mod1  mod2
1  .30      2        wai
2   0.1     4       other
2   0.2      4        calpas
3   0.1     1         itas
3   0.25    1         wai

Any help is much appreciated!

AC Del Re

   [[alternative HTML version deleted]]

Thank you, Dennis and Petr.

One more question:  when aggregating to one es per id, how would I go 

keeping the other variables in the data.frame (e.g., keeping the value 

the first row of the other variables, such as mod2) e.g.:

# Dennis provided this example (notice how mod2 is removed from the 

with(x, aggregate(list(es = es), by = list(id = id, mod1 = mod1), 

  id mod1   es
1  3    1 0.20
2  1    2 0.30
3  2    4 0.15

# How can I get this output (taking the first row of the other variable 

the data.frame):

id  es   mod1  mod2

1  .30     2        wai
2  .15     4        other
3  .20     1         itas


Thank you,

AC


On Thu, Jan 28, 2010 at 1:29 AM, Petr PIKAL <petr.pikal at precheza.cz> 

HI

r-help-bounces at r-project.org napsal dne 28.01.2010 04:35:29:

Hi All,

I'm conducting a meta-analysis and have taken a data.frame with

multiple

rows per
study (for each effect size) and performed a weighted average of

effect

size for
each study. This results in a reduced # of rows. I am particularly
interested in
simply reducing the additional variables in the data.frame to the

first row

of the
corresponding id variable. For example:

id<-c(1,2,2,3,3,3)
es<-c(.3,.1,.3,.1,.2,.3)
mod1<-c(2,4,4,1,1,1)
mod2<-c("wai","other","calpas","wai","itas","other")
data<-as.data.frame(cbind(id,es,mod1,mod2))

Do not use cbind. Its output is a matrix and in this case character
matrix. Resulting data frame will consist from factors as you can 

str(data)

data<-data.frame(id=id,es=es,mod1=mod1,mod2=mod2)

data

   id   es    mod1 mod2
1  1   0.3    2     wai
2  2   0.1    4     other
3  2   0.2    4     calpas
4  3   0.1    1     itas
5  3   0.2    1     wai
6  3   0.3    1     wai

# I would like to reduce the entire data.frame like this:

E.g. aggregate

aggregate(data[, -(3:4)], data[,3:4], mean)
 mod1   mod2 id  es
1    4 calpas  2 0.3
2    1   itas  3 0.2
3    1  other  3 0.3
4    4  other  2 0.1
5    1    wai  3 0.1
6    2    wai  1 0.3

doBy or tapply or ddply from plyr library or ....

Regards
Petr

id  es   mod1  mod2

1  .30     2        wai
2  .15     4        other
3  .20     1         itas

# If possible, I would also like the option of this (collapsing on 

and

mod2):

id  es   mod1  mod2
1  .30      2        wai
2   0.1     4       other
2   0.2      4        calpas
3   0.1     1         itas
3   0.25    1         wai

Any help is much appreciated!

AC Del Re

   [[alternative HTML version deleted]]

   [[alternative HTML version deleted]]