Finding the minimums of groups of observations in R - r

Question

I'm relatively new to R and struggle with "vectorizing" all my code in R. Even though I appreciate that's the proper way to do it.
I need to set a value in a dataframe to be the minimum time for the IDs.
ID isTrue RealTime MinTime
1 TRUE 16
1 FALSE 8
1 TRUE 10
2 TRUE 7
2 TRUE 30
3 FALSE 3
To be turned into:
ID isTrue RealTime MinTime
1 TRUE 16 10
1 FALSE 8
1 TRUE 10 10
2 TRUE 7 7
2 TRUE 30 7
3 FALSE 3
The following works perfectly. However, it takes 10 minutes to run which isn't ideal:
for (i in 1:nrow(df)){
if (df[i,'isTrue']) {
prevTime <- sqldf(paste('Select min(MinTime) from dfStageIV where ID =',df[i,'ID'],sep=" "))[1,1]
if (is.na(prevTime) | is.na(df[i,'MinTime']) | df[i,'MinTime'] < prevTime){
df[i,'MinTime']<-dfStageIV[i,'RealTime']
} else {
dfStageIV[i,'MinTime']<-prevTime
}
}
}
How should I do this properly? I take it using for or do loops are not the best way in R. I've been looking at the apply() and aggregate.data.frame() functions but can't make sense of how to do this. Can someone point me in the right direction? Much appreciated!!

Answer 1

Here is a two line base R solution using ave, pmax, and is.na.
# calculate minimum for each ID, excluding FALSE instances
df$MinTime <- ave(pmax(df$RealTime, (!df$isTrue) * max(df$RealTime)), df$ID, FUN=min)
# turn FALSE instances into NA
is.na(df$MinTime) <- (!df$isTrue)
which returns
df
ID isTrue RealTime MinTime
1 1 TRUE 16 10
2 1 FALSE 8 NA
3 1 TRUE 10 10
4 2 TRUE 7 7
5 2 TRUE 30 7
6 3 FALSE 3 NA
In the first line, pmax is used to construct a vector of the observations if df$isTrue is TRUE or the maximum RealTime value in the data.frame. This new vector is used in the minimum calculation. The FALSE values are set to NA in the second line.
data
df <- read.table(header=T, text="ID isTrue RealTime
1 TRUE 16
1 FALSE 8
1 TRUE 10
2 TRUE 7
2 TRUE 30
3 FALSE 3")

Answer 2

It should be far faster with a dplyr chain. Here we group the data frame by both ID and group and get the minima at the group level. Then we can ungroup it again and simply remove the F minima.
library(dplyr)
df %>%
group_by(ID, isTrue) %>%
mutate(Min.all = min(RealTime)) %>%
ungroup() %>%
transmute(ID, isTrue, RealTime, MinTime = ifelse(isTrue == T, Min.all, ""))
Output:
# A tibble: 6 × 4
ID isTrue RealTime MinTime
<int> <lgl> <int> <chr>
1 1 TRUE 16 10
2 1 FALSE 8
3 1 TRUE 10 10
4 2 TRUE 7 7
5 2 TRUE 30 7
6 3 FALSE 3
I'd really recommend you get familiar with dplyr if you're going to be doing lots of data frame manipulation.

Answer 3

Someone suggested using the ave() function and the following works and is fast although it returns a ton of warnings:
df$MinTime<-ave(df$RealTime,df$ID, df$isTrue, FUN = min)
df$MinTime<-ifelse(df$isTrue, df$MinTime,NA).

Answer 4

The code in the question could be simplified by doing it all in SQL or all in R (appropriately vectorized) rather than half and half. There are already some R solutions so here is an SQL solution that shows that the problem amounts to aggregating a custom self-join.
library(sqldf)
sqldf("select a.*, min(b.RealTime) minRealTime
from df a
left join df b on a.ID = b.ID and a.isTRUE and b.isTRUE
group by a.rowid")
giving:
ID isTrue RealTime minRealTime
1 1 TRUE 16 10
2 1 FALSE 8 NA
3 1 TRUE 10 10
4 2 TRUE 7 7
5 2 TRUE 30 7
6 3 FALSE 3 NA