remove rows based on substraction results - r

Question

I have a large data set like this:
df <- data.frame(group = c(rep(1, 6), rep(5, 6)), score = c(30, 10, 22, 44, 6, 5, 20, 35, 2, 60, 14,5))
group score
1 1 30
2 1 10
3 1 22
4 1 44
5 1 6
6 1 5
7 5 20
8 5 35
9 5 2
10 5 60
11 5 14
12 5 5
...
I want to do a subtraction for each neighboring score within each group, if the difference is greater than 30, remove the smaller score. For example, within group 1, 30-10=20<30, 10-22=-12<30, 22-44=-22<30, 44-6=38>30 (remove 6), 44-5=39>30 (remove 5)... The expected output should look like this:
group score
1 1 30
2 1 10
3 1 22
4 1 44
5 5 20
6 5 35
7 5 60
...
Does anyone have idea about realizing this?

Answer 1

Like this?
repeat {
df$diff=unlist(by(df$score,df$group,function(x)c(0,-diff(x))))
if (all(df$diff<30)) break
df <- df[df$diff<30,]
}
df$diff <- NULL
df
# group score
# 1 1 30
# 2 1 10
# 3 1 22
# 4 1 44
# 7 5 20
# 8 5 35
# 10 5 60
This (seems...) to require an iterative approach, because the "neighboring score" changes after removal of a row. So before you remove 6, the difference 44 - 6 > 30, but 6 - 5 < 30. After you remove 6, the difference 44 - 5 > 30.
So this calculates difference between successive rows by group (using by(...) and diff(...)), and removes the appropriate rows, then repeats the process until all differences are < 30.

Answer 2

It's not elegant but it should work:
out = data.frame(group = numeric(), score=numeric())
#cycle through the groups
for(g in levels(as.factor(df$group))){
temp = subset(df, df$group==g)
#now go through the scores
left = temp$score[1]
for(s in seq(2, length(temp$score))){
if(left - temp$score[s] > 30){#Test the condition
temp$score[s] = NA
}else{
left = temp$score[s] #if condition not met then the
}
}
#Add only the rows without NAs to the out
out = rbind(out, temp[which(!is.na(temp$score)),])
}
There should be a way to do this using ave but carrying the last value when removing the next if the diff >30 is tricky! I'd appreciate the more elegant solution if there is one.

Answer 3

You can try
df
## group score
## 1 1 30
## 2 1 10
## 3 1 22
## 4 1 44
## 5 1 6
## 6 1 5
## 7 5 20
## 8 5 35
## 9 5 2
## 10 5 60
## 11 5 14
## 12 5 5
tmp <- df[!unlist(tapply(df$score, df$group, FUN = function(x) c(F, -diff(x) > 30), simplify = T)), ]
while (!identical(df, tmp)) {
df <- tmp
tmp <- df[!unlist(tapply(df$score, df$group, FUN = function(x) c(F, -diff(x) > 30), simplify = T)), ]
}
tmp
## group score
## 1 1 30
## 2 1 10
## 3 1 22
## 4 1 44
## 7 5 20
## 8 5 35
## 10 5 60