0

I would like too summarize one column by another, which I know how to do, using either ddply or aggregate...But, I want to also add groups within my grouping variable to summarize the data in a custom fashion, see below:

##Create Data
year<-c(2008,2008,2008,2008,2009,2009,2009,2010,2010,2010,2010,2011,2011,2011,2011,2011)
catch<-c(1,2,0,1,1,0,1,0,1,2,1,0,1,1,1,0)
df<-data.frame(year,catch)

##I know how to aggregate by year:
yearlyAVcap<-ddply(df,~year,summarise,mean=mean(catch),sd=sd(catch))

##But I'd like to have 2008,2009,2010 summarized together and 2011 on it's own (i,e two groups-    1=2008,2009,2010 ; 2=2011)
ctlamb
  • 131
  • 1
  • 4
  • 14

1 Answers1

1

You could try

df1 <- transform(df, yrGroup=c('2008-2010', '2011')[(year==2011)+1])


library(plyr)
ddply(df1, .(yrGroup), summarise, mean=mean(catch), sd=sd(catch))
#  yrGroup        mean        sd
#1 2008-2010 0.9090909 0.7006490
#2      2011 0.6000000 0.5477226

Or using dplyr

 df1 %>%
      group_by(yrGroup) %>%
      summarise_each(funs(mean, sd), catch)
 #summarise_each can be used for multiple columns 
#    yrGroup      mean        sd
#1 2008-2010 0.9090909 0.7006490
#2      2011 0.6000000 0.5477226

Explanation

To get two groups from different years, in the example (2008-2010 and 2011), we can use == to match for the second group year i.e. 2011. The result will be a logical vector. Then, we can transform this to numeric index by adding 1 to it so that the two groups will 1 and 2. If we add 0, the TRUE will be converted to 1 and FALSE as 0. This numeric index can be used for renaming it to 2008-2010 and 2011. 

df$year==2011
 #[1] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE  TRUE
#[13]  TRUE  TRUE  TRUE  TRUE

(df$year==2011)+1
 #[1] 1 1 1 1 1 1 1 1 1 1 1 2 2 2 2 2

 c('2008-2010', '2011')[(df$year==2011)+1]
 #[1] "2008-2010" "2008-2010" "2008-2010" "2008-2010" "2008-2010" "2008-2010"
 #[7] "2008-2010" "2008-2010" "2008-2010" "2008-2010" "2008-2010" "2011"     
 #[13] "2011"      "2011"      "2011"      "2011"

Suppose, we have multiple years, for example 2008-2010 as one group and 2011-2013 as another group, we can use %in%

  (df$year %in% 2011:2013)+1
akrun
  • 874,273
  • 37
  • 540
  • 662
  • Thanks!, I understand what the yrGroup=c('2008-2010', '2011') portion does, but what does this [(year==2011)+1]) do? – ctlamb Nov 08 '14 at 16:18
  • @ctlamb `year==2001` returns a logical vector. If you add `1` to add, you get a numeric index of `1 and 2` – akrun Nov 08 '14 at 16:21
  • And if I had more years, for example 2012 and 2013, that I wanted to be added, would I do [((year==2011)+1)((year==2012)+1)] etc? – ctlamb Nov 08 '14 at 16:22
  • @ctlamb It depends upon the groups you wanted. Here, in the question, you asked for `2011` to be a single group. If you need `2012` as a single group, yes, you can replace it with `year==2012` – akrun Nov 08 '14 at 16:26