I have question about ddply and subset.
I have dataframe df like this :
df <- read.table(textConnection(
" id v_idn v_seed v_time v_pop v_rank v_perco
1 15 125648 0 150 1 15
2 17 125648 0 120 2 5
3 18 125648开发者_Python百科 0 100 3 6
4 52 125648 0 25 4 1
5 17 125648 10 220 1 5
6 15 125648 10 160 2 15
7 18 125648 10 110 3 6
8 52 125648 10 50 4 1
9 56 -11152 0 250 1 17
10 15 -11152 0 180 2 15
11 18 -11152 0 110 3 6
12 22 -11152 0 5 4 14
13 56 -11152 10 250 1 17
14 15 -11152 10 180 2 15
15 22 -11152 10 125 3 14
16 18 -11152 10 120 4 6 "), header=TRUE)
STEP ONE :
I have a list of equal interval with cut_interval like this :
myinterval <- cut_interval(c(15,5,6,1,17,14), length=10)
So i have two levels here : [0,10) and (10,20]
STEP TWO :
I want each group/class is define by my two levels in v_cut ... like this :
id v_idn v_seed v_time v_pop v_rank v_perco v_cut
1 15 125648 0 150 1 15 (10,20]
2 17 125648 0 120 2 5 [0,10)
3 18 125648 0 100 3 6 [0,10)
4 52 125648 0 25 4 1 [0,10)
5 17 125648 10 220 1 5 [0,10)
6 15 125648 10 160 2 15 (10,20]
7 18 125648 10 110 3 6 [0,10)
8 52 125648 10 50 4 1 [0,10)
9 56 -11152 0 250 1 17 (10,20]
10 15 -11152 0 180 2 15 (10,20]
11 18 -11152 0 110 3 6 [0,10)
12 22 -11152 0 5 4 14 (10,20]
13 56 -11152 10 250 1 17 (10,20]
14 15 -11152 10 180 2 15 (10,20]
15 22 -11152 10 125 3 14 (10,20]
16 18 -11152 10 120 4 6 [0,10)
STEP 3 :
I want to know the variability of v_rank for x axis, and time for y axis, for each group v_cut, so i need to compute min,mean,max,sd for v_rank value with something like
ddply(df, .(v_cut,v_time), summarize ,mean = mean(v_rank), min = min(v_rank), max = max(v_rank), sd = sd(v_rank))
*RESULT WANTED : *
id v_time MEAN.v_rank ... v_cut
1 0 2.25 (10,20]
2 0 2.42 [0,10)
3 10 2.25 [0,10)
4 10 2.42 (10,20]
MY PROBLEM
I don't know how to pass step 1 -> step 2 :/
And if it's possible to group by v_cut like my example in step 3 ?
Is there a possibility to make the same things with the "subset" option of ddply ?
One more time, thanks a lot for your help great R guru !
UPDATE 1 :
I have an answer to go step1 to step2 :
df$v_cut <- cut_interval(df$v_perco,n=10)
I'm using plyr, but there are perhaps a better answer in this case ?
Answer to go to step 2 to step 3 ?
UPDATE 2 :
Brandon Bertelsen give me a good answer with melt + cast, but now (to understand) i want to make the same operation with plyr and ddply .. with a different result :
id v_idn v_time MEAN.v_rank ... v_cut
1 15 0 2.25 (10,20]
2 15 10 2.45 (10,20]
2 17 0 1.52 [0,10)
2 17 10 2.42 [0,10)
etc.
I'm trying with something like this :
r('sumData <- ddply(df, .(v_idn,v_time), summarize,min = min(v_rank),mean = mean(v_rank), max = max(v_rank), sd=sd(v_rank))')
But i want to have v_cut in my sumData dataframe, how can i do with ddply ? is there an option to make this ? Or merging with initial df and key = v_idn to add column v_cut to sumData is the only good answer ?
You don't really need plyr for this, you can use reshape
## Pull what you need
dfx <- df[c("v_seed", "v_time","v_rank","v_perco")]
## Bring in your cuts
dfx <- data.frame(dfx, ifelse(df$v_perco > 10,"(10,20]", "[0,10)")))
## Rename v_cut
colnames(dfx)[ncol(dfx)] <- "v_cut"
## Melt it.
dfx <- melt(dfx, id=c("v_cut", "v_seed", "v_time"))
## Cast it.
dfx <- cast(dfx, v_cut + v_time + v_seed ~ variable, c(mean,min,max,sd))
if you only want the mean, then replace the last line with:
dfx <- cast(dfx, v_cut + v_time + v_seed ~ variable, mean)
type "dfx" and you'll see a data frame with what you asked for.
You're just having a problem with syntax is all:
## Add your cut
df.new <- data.frame(df, ifelse(df$v_perco > 10,"(10,20]", "[0,10)"))
## Rename v_cut
colnames(df.new)[ncol(df.new)] <- "v_cut"
## Careful here read the note below
df.new <- ddply(df.new, .(v_idn, v_time), function(x) unique(data.frame(
mean = mean(x$v_rank),
v_cut = x$v_cut
)))
Alternatively:
ddply(df.new, .(v_idn, v_time), summarise, mean=mean(v_rank))
With ".(v_idn, v_time)" you're telling ddply that for each combination of v_idn and v_time, you want it to calculate the mean of v_rank.
精彩评论