Using 'window' functions in dplyr - r

I need to process rows of a data-frame in order, but need to look-back for certain rows. Here is an approximate example:
library(dplyr)
d <- data_frame(trial = rep(c("A","a","b","B","x","y"),2))
d <- d %>%
mutate(cond = rep('', n()), num = as.integer(rep(0,n())))
for (i in 1:nrow(d)){
if(d$trial[i] == "A"){
d$num[i] <- 0
d$cond[i] <- "A"
}
else if(d$trial[i] == "B"){
d$num[i] <- 0
d$cond[i] <- "B"
}
else{
d$num[i] <- d$num[i-1] +1
d$cond[i] <- d$cond[i-1]
}
}
The resulting data-frame looks like
> d
Source: local data frame [12 x 3]
trial cond num
1 A A 0
2 a A 1
3 b A 2
4 B B 0
5 x B 1
6 y B 2
7 A A 0
8 a A 1
9 b A 2
10 B B 0
11 x B 1
12 y B 2
What is the proper way of doing this using dplyr?

dlpyr-only solution:
d %>%
group_by(i=cumsum(trial %in% c('A','B'))) %>%
mutate(cond=trial[1],num=seq(n())-1) %>%
ungroup() %>%
select(-i)
# trial cond num
# 1 A A 0
# 2 a A 1
# 3 b A 2
# 4 B B 0
# 5 x B 1
# 6 y B 2
# 7 A A 0
# 8 a A 1
# 9 b A 2
# 10 B B 0
# 11 x B 1
# 12 y B 2

Try
d %>%
mutate(cond = zoo::na.locf(ifelse(trial=="A"|trial=="B", trial, NA))) %>%
group_by(id=rep(1:length(rle(cond)$values), rle(cond)$lengths)) %>%
mutate(num = 0:(n()-1)) %>% ungroup %>%
select(-id)

Here is one way. The first thing was to add A or B in cond using ifelse. Then, I employed na.locf() from the zoo package in order to fill NA with A or B. I wanted to assign a temporary group ID before I took care of num. I borrowed rleid() in the data.table package. Grouping the data with the temporary group ID (i.e., foo), I used row_number() which is one of the window functions in the dplyr package. Note that I tried to remove foo doing select(-foo). But, the column wanted to stay. I think this is probably something to do with compatibility of the function.
library(zoo)
library(dplyr)
library(data.table)
d <- data_frame(trial = rep(c("A","a","b","B","x","y"),2))
mutate(d, cond = ifelse(trial == "A" | trial == "B", trial, NA),
cond = na.locf(cond),
foo = rleid(cond)) %>%
group_by(foo) %>%
mutate(num = row_number() - 1)
# trial cond foo num
#1 A A 1 0
#2 a A 1 1
#3 b A 1 2
#4 B B 2 0
#5 x B 2 1
#6 y B 2 2
#7 A A 3 0
#8 a A 3 1
#9 b A 3 2
#10 B B 4 0
#11 x B 4 1
#12 y B 4 2

Related

Find 2 out of 3 conditions per ID

I have the following dataframe:
df <-read.table(header=TRUE, text="id code
1 A
1 B
1 C
2 A
2 A
2 A
3 A
3 B
3 A")
Per id, I would love to find those individuals that have at least 2 conditions, namely:
conditionA = "A"
conditionB = "B"
conditionC = "C"
and create a new colum with "index", 1 if there are two or more conditions met and 0 otherwise:
df_output <-read.table(header=TRUE, text="id code index
1 A 1
1 B 1
1 C 1
2 A 0
2 A 0
2 A 0
3 A 1
3 B 1
3 A 1")
So far I have tried the following:
df_output = df %>%
group_by(id) %>%
mutate(index = ifelse(grepl(conditionA|conditionB|conditionC, code), 1, 0))
and as you can see I am struggling to get the threshold count into the code.
You can create a vector of conditions, and then use %in% and sum to count the number of occurrences in each group. Use + (or ifelse) to convert logical into 1 and 0:
conditions = c("A", "B", "C")
df %>%
group_by(id) %>%
mutate(index = +(sum(unique(code) %in% conditions) >= 2))
id code index
1 1 A 1
2 1 B 1
3 1 C 1
4 2 A 0
5 2 A 0
6 2 A 0
7 3 A 1
8 3 B 1
9 3 A 1
You could use n_distinct(), which is a faster and more concise equivalent of length(unique(x)).
df %>%
group_by(id) %>%
mutate(index = +(n_distinct(code) >= 2)) %>%
ungroup()
# # A tibble: 9 × 3
# id code index
# <int> <chr> <int>
# 1 1 A 1
# 2 1 B 1
# 3 1 C 1
# 4 2 A 0
# 5 2 A 0
# 6 2 A 0
# 7 3 A 1
# 8 3 B 1
# 9 3 A 1
You can check conditions using intersect() function and check whether resulting list is of minimal (eg- 2) length.
conditions = c('A', 'B', 'C')
df_output2 =
df %>%
group_by(id) %>%
mutate(index = as.integer(length(intersect(code, conditions)) >= 2))

Filter (subset) by conditions in 2 columns in R (dplyr or otherwise)

Given a dataset such as:
set.seed(134)
df<- data.frame(ID= rep(LETTERS[1:5], each=2),
condition=rep(0:1, 5),
value=rpois(10, 3)
)
df
ID condition value
1 A 0 2
2 A 1 3
3 B 0 5
4 B 1 2
5 C 0 3
6 C 1 1
7 D 0 2
8 D 1 4
9 E 0 1
10 E 1 5
For each ID, when the value for condition==0 is less than the value for condition==1, I want to keep both observations. When the value for condition==0 is greater than condition==1, I want to keep only the row for condition==0.
The subset returned should be this:
ID condition value
1 A 0 2
2 A 1 3
3 B 0 5
5 C 0 3
7 D 0 2
8 D 1 4
9 E 0 1
10 E 1 5
Using dplyr the first step is:
df %>% group_by(ID) %>%
But not sure where to go from there.
Translating fairly literally,
library(dplyr)
set.seed(134)
df <- data.frame(ID = rep(LETTERS[1:5], each = 2),
condition = rep(0:1, 5),
value = rpois(10, 3))
df %>% group_by(ID) %>%
filter(condition == 0 |
(condition == 1 & value > value[condition == 0]))
#> # A tibble: 8 x 3
#> # Groups: ID [5]
#> ID condition value
#> <fct> <int> <int>
#> 1 A 0 2
#> 2 A 1 3
#> 3 B 0 5
#> 4 C 0 3
#> 5 D 0 2
#> 6 D 1 4
#> 7 E 0 1
#> 8 E 1 5
This depends on each group having a single observation with condition == 0, but should otherwise be fairly robust.
This is may not be the easiest way, but should work as you want.
library(reshape2)
df %>%
dcast(ID ~ condition, value.var = 'value') %>% # cast to wide format
mutate(`1` = ifelse(`1` > `0`, `1`, NA)) %>% # turn 0>1 values as NA
melt('ID') %>% # melt as long format
arrange(ID) %>% # sort by ID
filter(complete.cases(.)) # remove NA rows
Output:
ID variable value
1 A 0 2
2 A 1 3
3 B 0 5
4 C 0 3
5 D 0 2
6 D 1 4
7 E 0 1
8 E 1 5
You always want the value from the first row in each group. You only want the value from the second row in each group if it's larger than the first.
This works:
df %>%
group_by(ID) %>%
filter(row_number() == 1 | value > lag(value))
Edit: as #alistaire points out, this method depends on a particular order in, which is might be a good idea to guarantee as follows:
df %>%
arrange(ID, condition) %>%
group_by(ID) %>%
filter(row_number() == 1 | value > lag(value))

Extracting group dependent results from a dataframe

I have a dataframe made from different groups, and for each group real and predicted values. I want to extract values of tests on these values :
library(dplyr)
d = data.frame(group = c(rep(5,x="a"),rep(5,x="b")), real = c(rep(2, x=1:5)), pred = c(2,1,3,4,5,1,2,4,3,5))
group real pred
1 a 1 2
2 a 2 1
3 a 3 3
4 a 4 4
5 a 5 5
6 b 1 1
7 b 2 2
8 b 3 4
9 b 4 3
10 b 5 5
d <- d %>% group_by(group) %>% mutate( sg = ifelse(real == 1 & real == pred, 1, 0))
d <- d %>% group_by(group) %>% mutate( sp = ifelse(real <= 3 & pred <= 3, 1, 0))
d %>% distinct(sg, sp)
sg sp group
1 0 1 a
2 0 0 a
3 1 1 b
4 0 1 b
5 0 0 b
But I want something like this (only 1 result per group)
sg sp group
1 0 1 a
3 1 1 b
I am pretty sure dplyr, data.table or tidyr can do something but I cannot find how.
If it is always the first row of each group that you want to extract, you could use the do function:
d %>% do(.[1,])
Another option is to use the filter function like this:
d %>% filter(seq_along(sp) == 1)

group by and filter data management using dplyr

Take a simple dataset
a <- c(1,2,3,4,5,6,7,8)
b <- c(1,2,2,1,2,2,2,2)
c <- c(1,1,1,2,2,2,3,3)
d <- data.frame(a,b,c)
now I want to filter my data, so that we group_by(c) and then remove all data where no b=1occurs.
Thus the results (e) should look like d but without the two bottom rows
I have tried using
e <- d %>%
group_by(c) %>%
filter(n(b)>1)
The output should contain the data in green below and remove the data in red
Try
d %>%
group_by(c) %>%
filter(any(b == 1))
Which gives:
#Source: local data frame [6 x 3]
#Groups: c
#
# a b c
#1 1 1 1
#2 2 2 1
#3 3 2 1
#4 4 1 2
#5 5 2 2
#6 6 2 2
You can try
df <- d %>% mutate(test = ifelse((b != 1) == T, 0, 1)) %>% group_by(c) %>%
mutate(test = sum(test)) %>% filter(test != 0) %>% select(-test)
which yields
# a b c
#1 1 1 1
#2 2 2 1
#3 3 2 1
#4 4 1 2
#5 5 2 2
#6 6 2 2

R, dplyr: cumulative version of n_distinct

I have a dataframe as follows. It is ordered by column time.
Input -
df = data.frame(time = 1:20,
grp = sort(rep(1:5,4)),
var1 = rep(c('A','B'),10)
)
head(df,10)
time grp var1
1 1 1 A
2 2 1 B
3 3 1 A
4 4 1 B
5 5 2 A
6 6 2 B
7 7 2 A
8 8 2 B
9 9 3 A
10 10 3 B
I want to create another variable var2 which computes no of distinct var1 values so far i.e. until that point in time for each group grp . This is a little different from what I'd get if I were to use n_distinct.
Expected output -
time grp var1 var2
1 1 1 A 1
2 2 1 B 2
3 3 1 A 2
4 4 1 B 2
5 5 2 A 1
6 6 2 B 2
7 7 2 A 2
8 8 2 B 2
9 9 3 A 1
10 10 3 B 2
I want to create a function say cum_n_distinct for this and use it as -
d_out = df %>%
arrange(time) %>%
group_by(grp) %>%
mutate(var2 = cum_n_distinct(var1))
A dplyr solution inspired from #akrun's answer -
Ths logic is basically to set 1st occurrence of each unique values of var1 to 1 and rest to 0 for each group grp and then apply cumsum on it -
df = df %>%
arrange(time) %>%
group_by(grp,var1) %>%
mutate(var_temp = ifelse(row_number()==1,1,0)) %>%
group_by(grp) %>%
mutate(var2 = cumsum(var_temp)) %>%
select(-var_temp)
head(df,10)
Source: local data frame [10 x 4]
Groups: grp
time grp var1 var2
1 1 1 A 1
2 2 1 B 2
3 3 1 A 2
4 4 1 B 2
5 5 2 A 1
6 6 2 B 2
7 7 2 A 2
8 8 2 B 2
9 9 3 A 1
10 10 3 B 2
Assuming stuff is ordered by time already, first define a cumulative distinct function:
dist_cum <- function(var)
sapply(seq_along(var), function(x) length(unique(head(var, x))))
Then a base solution that uses ave to create groups (note, assumes var1 is factor), and then applies our function to each group:
transform(df, var2=ave(as.integer(var1), grp, FUN=dist_cum))
A data.table solution, basically doing the same thing:
library(data.table)
(data.table(df)[, var2:=dist_cum(var1), by=grp])
And dplyr, again, same thing:
library(dplyr)
df %>% group_by(grp) %>% mutate(var2=dist_cum(var1))
Try:
Update
With your new dataset, an approach in base R
df$var2 <- unlist(lapply(split(df, df$grp),
function(x) {x$var2 <-0
indx <- match(unique(x$var1), x$var1)
x$var2[indx] <- 1
cumsum(x$var2) }))
head(df,7)
# time grp var1 var2
# 1 1 1 A 1
# 2 2 1 B 2
# 3 3 1 A 2
# 4 4 1 B 2
# 5 5 2 A 1
# 6 6 2 B 2
# 7 7 2 A 2
Here's another solution using data.table that's pretty quick.
Generic Function
cum_n_distinct <- function(x, na.include = TRUE){
# Given a vector x, returns a corresponding vector y
# where the ith element of y gives the number of unique
# elements observed up to and including index i
# if na.include = TRUE (default) NA is counted as an
# additional unique element, otherwise it's essentially ignored
temp <- data.table(x, idx = seq_along(x))
firsts <- temp[temp[, .I[1L], by = x]$V1]
if(na.include == FALSE) firsts <- firsts[!is.na(x)]
y <- rep(0, times = length(x))
y[firsts$idx] <- 1
y <- cumsum(y)
return(y)
}
Example Use
cum_n_distinct(c(5,10,10,15,5)) # 1 2 2 3 3
cum_n_distinct(c(5,NA,10,15,5)) # 1 2 3 4 4
cum_n_distinct(c(5,NA,10,15,5), na.include = FALSE) # 1 1 2 3 3
Solution To Your Question
d_out = df %>%
arrange(time) %>%
group_by(grp) %>%
mutate(var2 = cum_n_distinct(var1))

Resources