The following code generates 2-level groups (by State within Test), then ranks each observation within each group based on ascending order of Grade. School is the tie-breaker.
School<-rep(c("A","B","C","D"),each=10)
State<-rep(c("NY","NJ"),times=20)
Test<-rep(c("LSAT", "MCAT", "GRE","TOEFL","ACT"), times=8)
Grade<-trunc(rep((seq(from=500, to=600,length.out=4))))
dat<-data.frame(Test,State,School,Grade)
library(plyr)
dat<-ddply(dat, .(Test, State),transform,num=rank(Grade,ties.method="first"))
I convert the first-ranked item within each group to "lowest" using the following code:
dat$num[dat$num==1]<-"lowest"
In this sample df, the number of items per group is always 4, so I can convert the highest-ranked item in each group to "highest" using the following code:
dat$num[dat$num==4]<-"highest"
But how can I tag observations with "highest" when the number of rows is not constant across all groups? The following code creates a version of the df with two additional rows in one of the groups.
School<-rep(c("A","B","C","D"),each=10)
State<-rep(c("NY","NJ"),times=20)
Test<-rep(c("LSAT", "MCAT", "GRE","TOEFL","ACT"), times=8)
Grade<-trunc(rep((seq(from=500, to=600,length.out=4))))
dat1<-data.frame(Test,State,School,Grade)
dat1<-rbind(dat1,
data.frame(Test="ACT",State="NJ",School="E",Grade=550),
data.frame(Test="ACT",State="NJ",School="F",Grade=650))
library(plyr)
dat1<-ddply(dat1, .(Test, State),transform,num=rank(Grade,ties.method="first"))
You can do that by checking which is the highest/lowest in each group and assign highest/lowest to those rows. Here, I use ddply to do that since you already use plyr in your code:
dat1 <- ddply(dat1, .(Test, State), transform, num=ifelse(num == max(num), "highest",
ifelse(num == min(num), "lowest", num)))
> dat1
Test State School Grade num
1 ACT NJ A 533 lowest
2 ACT NJ B 600 4
3 ACT NJ C 533 2
4 ACT NJ D 600 5
5 ACT NJ E 550 3
6 ACT NJ F 650 highest
7 ACT NY A 500 lowest
8 ACT NY B 566 3
9 ACT NY C 500 2
10 ACT NY D 566 highest
11 GRE NJ A 600 3
12 GRE NJ B 533 lowest
13 GRE NJ C 600 highest
14 GRE NJ D 533 2
15 GRE NY A 566 3
16 GRE NY B 500 lowest
17 GRE NY C 566 highest
18 GRE NY D 500 2
19 LSAT NJ A 533 lowest
20 LSAT NJ B 600 3
21 LSAT NJ C 533 2
22 LSAT NJ D 600 highest
23 LSAT NY A 500 lowest
24 LSAT NY B 566 3
25 LSAT NY C 500 2
26 LSAT NY D 566 highest
27 MCAT NJ A 533 lowest
28 MCAT NJ B 600 3
29 MCAT NJ C 533 2
30 MCAT NJ D 600 highest
31 MCAT NY A 566 3
32 MCAT NY B 500 lowest
33 MCAT NY C 566 highest
34 MCAT NY D 500 2
35 TOEFL NJ A 600 3
36 TOEFL NJ B 533 lowest
37 TOEFL NJ C 600 highest
38 TOEFL NJ D 533 2
39 TOEFL NY A 500 lowest
40 TOEFL NY B 566 3
41 TOEFL NY C 500 2
42 TOEFL NY D 566 highest
If your data is large enough, you could also consider using dplyr or data.table which will be faster than plyr.