R - Creating and updating a reference value in a for loop - r

Question

I have the following (simulated) dataset
m=500
n=8
df<-data.frame(matrix(sample(0:1,m*n, replace=TRUE),m,n))
df$ID<-c(1:20)
attach(df)
df<-df[order(ID),]
df$round<-c(1:25)
df$payoff<-runif(n=500, min=1e-12, max=.9999999999)
First, I want a for loop that allows me to compare each row with the one before, so that the output takes value 1 if the payoff of the row is greater than the payoff of the row before. Then, I want the row with the highest payoff that was found so far to function as a reference for the next rows, so that the output takes now value 1 if the payoff of the next row is grater than the payoff of the row with the highest payoff that was found so far. The reference needs to be progressively updated as soon as a new highest value is found.
I managed to build a loop for the first step
df_split <- split(df, df$ID)
y<-data.frame("ID"=NULL, "round"=NULL, "feedback"=NULL)
for (i in 1:length(df_split)) {
myvector<-as.matrix(df_split[[i]][-1:-10])
for (j in 2:nrow(myvector)){
feedb<-ifelse(myvector[j,] > myvector[j-1,], 1, 0)
df2<-data.frame("ID"=i, "round"=j, "feedback"=feedb)
y<-rbind(y,df2)
}
}
Now I want to add to the loop the second step, that is indicating the row with the highest payoff that was found so far as reference, and compare the next row with such row. As already mentioned above, such a reference needs to be updated as a new highest value is found.
Does anybody have a solution?
Thank you for all your help!
EDIT:
Thank you both #r2evans and #Jon_Spring for you suggestions!
The reason why I am using a loop is that I need to calculate the output for each ID independently (sorry, I forgot to mention).
This is also why I am splitting the original dataframe into 20 dataframes (one per ID).
If I understand correctly your solutions, the codes are such that, when it comes for instance to ID = 2, the last payoff with the highest value is associated with ID = 1. The same happens when it comes to ID = 3, ID = 4, and so forth. Then, the resulting output is not correct, because the calculation should restart for each ID.
I didn't know the function cummax, thank you again! I'll try to integrate it into the logic of my loop, which also gives an output column as I need it.

Answer 1

I don't think you need any loops.
Up front, for reproducibility, I set my random seed with set.seed(1) before generating the frame above. This allows you to see the "exact same" frame as I'm creating below.
head(within(df, {
isbetter <- c(TRUE, diff(payoff) > 0)
maxsofar <- cummax(df$payoff)
maxsofar <- c(0, maxsofar[-length(maxsofar)])
isbestsofar <- as.integer(payoff > maxsofar)
}), n=20)
# X1 X2 X3 X4 X5 X6 X7 X8 ID round payoff isbestsofar maxsofar isbetter
# 1 0 1 1 0 1 1 1 1 1 1 0.18776846 1 0.0000000 TRUE
# 21 1 1 0 0 0 0 0 1 1 2 0.50475902 1 0.1877685 TRUE
# 41 1 0 0 0 0 1 0 0 1 3 0.02728685 0 0.5047590 FALSE
# 61 1 1 0 0 0 0 1 0 1 4 0.49629785 0 0.5047590 TRUE
# 81 0 0 0 0 1 1 1 0 1 5 0.94735171 1 0.5047590 TRUE
# 101 1 1 1 0 1 1 0 1 1 6 0.38118213 0 0.9473517 FALSE
# 121 1 1 0 1 0 0 1 0 1 7 0.69821373 0 0.9473517 TRUE
# 141 1 1 0 0 1 0 1 1 1 8 0.68876581 0 0.9473517 FALSE
# 161 0 0 0 0 1 0 0 0 1 9 0.47773068 0 0.9473517 FALSE
# 181 0 1 0 1 1 0 0 1 1 10 0.27334761 0 0.9473517 FALSE
# 201 0 1 0 1 1 0 1 0 1 11 0.75691633 0 0.9473517 TRUE
# 221 0 0 1 1 1 0 1 0 1 12 0.24753206 0 0.9473517 FALSE
# 241 0 0 0 1 0 1 1 0 1 13 0.52133948 0 0.9473517 TRUE
# 261 1 1 0 0 1 0 0 0 1 14 0.61284324 0 0.9473517 TRUE
# 281 0 1 0 1 1 0 1 0 1 15 0.09504998 0 0.9473517 FALSE
# 301 1 1 1 0 0 1 0 0 1 16 0.56575876 0 0.9473517 TRUE
# 321 1 0 1 1 0 1 1 1 1 17 0.01687416 0 0.9473517 FALSE
# 341 1 1 0 1 0 1 0 1 1 18 0.19987888 0 0.9473517 TRUE
# 361 0 0 1 1 1 0 0 1 1 19 0.41758380 0 0.9473517 TRUE
# 381 0 0 1 0 1 1 0 0 1 20 0.20550609 0 0.9473517 FALSE
I use within for simple creation/processing of columns within the data.frame; this could easily be done verbatim df$isbetter <- c(TRUE, diff(df$payoff) > 0), with dplyr, with data.table, or likely in other ways too. Take your pick, the logic and outcome should be effectively the same (other than column order, perhaps).

Answer 2

df$cummax = cummax(df$payoff)
df$new_max = df$payoff==df$cummax
Edit: added group_by, dplyr pipe
library(dplyr)
df2 <- df %>%
group_by(ID) %>%
mutate(cummax = cummax(payoff),
new_max = payoff==cummax) %>%
ungroup()
Output, showing what happens when we get to new ID:
> df2[20:30,]
# A tibble: 11 x 13
X1 X2 X3 X4 X5 X6 X7 X8 ID round payoff cummax new_max
<int> <int> <int> <int> <int> <int> <int> <int> <int> <int> <dbl> <dbl> <lgl>
1 0 0 1 0 1 1 0 0 1 20 0.206 0.947 FALSE
2 1 1 0 1 0 0 1 0 1 21 0.377 0.947 FALSE
3 0 0 1 0 0 0 1 0 1 22 0.0765 0.947 FALSE
4 0 0 1 1 0 0 0 0 1 23 0.145 0.947 FALSE
5 0 0 0 1 0 0 1 0 1 24 0.554 0.947 FALSE
6 1 0 0 0 1 1 1 1 1 25 0.662 0.947 FALSE
7 0 1 1 1 1 0 0 1 2 1 0.736 0.736 TRUE
8 0 1 1 1 1 0 0 0 2 2 0.376 0.736 FALSE
9 1 1 0 0 0 0 0 0 2 3 0.869 0.869 TRUE
10 0 0 1 1 1 0 1 1 2 4 0.795 0.869 FALSE
11 1 1 0 1 1 1 0 1 2 5 0.822 0.869 FALSE