I want to mix several observations in a data.frame using as a reference one constantly repeated variable.
Example:
id var1 var2 var3
a 1 na na
a na 2 na
a na na 3
b 1 na
b na 2 na
b na na na
c na na 3
c na 2 na
c 1 na na
Expected result:
id var1 var2 var3
a 1 2 3
b 1 2 na
c 1 2 3
A possible solution (replacing "na" by NA with na_if):
library(tidyverse)
df %>%
na_if("na") %>%
group_by(id) %>%
summarize(across(var1:var3, ~ sort(.x)[1]))
#> # A tibble: 3 × 4
#> id var1 var2 var3
#> <chr> <chr> <chr> <chr>
#> 1 a 1 2 3
#> 2 b 1 2 <NA>
#> 3 c 1 2 3
Assumptions:
"na" above is really R's native NA (not a string);
b's first row, var2 should be NA instead of an empty string ""
perhaps from the above, var1:var3 should be numbers
either you will never have a group where there is more than one non-NA in a group/column, or you don't care about anything other than the first and want the remaining discarded
library(dplyr)
dat %>%
group_by(id) %>%
summarize(across(everything(), ~ na.omit(.)[1]))
# # A tibble: 3 x 4
# id var1 var2 var3
# <chr> <int> <int> <int>
# 1 a 1 2 3
# 2 b 1 2 NA
# 3 c 1 2 3
Data
dat <- structure(list(id = c("a", "a", "a", "b", "b", "b", "c", "c", "c"), var1 = c(1L, NA, NA, 1L, NA, NA, NA, NA, 1L), var2 = c(NA, 2L, NA, NA, 2L, NA, NA, 2L, NA), var3 = c(NA, NA, 3L, NA, NA, NA, 3L, NA, NA)), class = "data.frame", row.names = c(NA, -9L))
Assuming that your data has NA, you can use the following base R option using the Data from #r2evans (thanks!):
aggregate(.~id, dat, mean, na.rm = TRUE, na.action=NULL)
Output:
id var1 var2 var3
1 a 1 2 3
2 b 1 2 NaN
3 c 1 2 3
Related
I have this type of data, where Sequis a grouping variable:
df <- data.frame(
Sequ = c(1,1,1,
2,2,2,
3,3,
4,4),
Answerer = c("A", NA, NA, "A", NA, NA, "B", NA, "C", NA),
PP_by = c(rep("A",5), rep("B",5)),
pp = c(0.1,0.2,0.3, 1, NA, NA, NA, NA, NA, NA)
)
I need to remove any Sequ where
(i) Answerer == PP_by AND
(ii) there is any NA in pp
I've tried this, but it obviously implements just the first condition (i):
library(dplyr)
df %>%
group_by(Sequ) %>%
filter(
all(!is.na(pp))
)
The expected result is:
Sequ Answerer PP_by pp
1 1 A A 0.1
2 1 <NA> A 0.2
3 1 <NA> A 0.3
9 4 C B NA
10 4 <NA> B NA
EDIT:
I've come up with this solution:
df %>%
group_by(Sequ) %>%
filter(
first(Answerer) != first(PP_by)
|
all(!is.na(pp))
)
Here's another way:
df %>%
group_by(Sequ) %>%
filter(!(
any(Answerer == PP_by, na.rm = TRUE) &
any(is.na(pp))
))
# # A tibble: 5 × 4
# # Groups: Sequ [2]
# Sequ Answerer PP_by pp
# <dbl> <chr> <chr> <dbl>
# 1 1 A A 0.1
# 2 1 NA A 0.2
# 3 1 NA A 0.3
# 4 4 C B NA
# 5 4 NA B NA
I have a large dataframe with approximately this pattern:
Person
Rate
Street
a
b
c
d
e
f
A
2
XYZ
1
NULL
3
4
5
NULL
A
2
XYZ
NULL
2
NULL
NULL
NULL
NULL
A
3
XYZ
NULL
NULL
NULL
NULL
NULL
6
B
2
DEF
NULL
NULL
NULL
NULL
5
NULL
B
2
DEF
NULL
2
3
NULL
NULL
6
C
1
DEF
1
2
3
4
5
6
A, b, c, d, e, f represents about 600 columns.
I am trying to combine the columns so that each person becomes one line, rows a-f combine into a single line using sum, and any conflicting rate or street information becomes a new row. So the data should look something like this:
Person
Rate
Rate 2
Street
a
b
c
d
e
f
A
2
3
XYZ
1
2
3
4
5
6
B
2
DEF
NULL
2
3
NULL
5
6
C
1
DEF
1
2
3
4
5
6
I keep trying to make this work with aggregate and summarize but I'm not sure that's the right approach.
Thank you very much for your help!
First we pivot all the unique rates per person and street.
library(reshape2)
tmp1=dcast(unique(df[,c("Person","Rate","Street")]),Person+Street~Rate,value.var="Rate")
colnames(tmp1)[-c(1:2)]=paste("Rate",colnames(tmp1)[-c(1:2)])
Then we aggregate and sum by person and rate, columns 4 to 9, from "a" to "f", change accordingly.
tmp2=aggregate(df[,4:9],list(Person=df$Person,Street=df$Street),function(x){
ifelse(all(is.na(x)),NA,sum(x,na.rm=T))
})
And finally merge the two.
merge(tmp1,tmp2,by=c("Person","Street"))
Person Street Rate 1 Rate 2 Rate 3 a b c d e f
1 A XYZ NA 2 3 1 2 3 4 5 6
2 B DEF NA 2 NA NA 2 3 NA 5 6
3 C DEF 1 NA NA 1 2 3 4 5 6
Perhaps, you can do this in two-step process -
library(dplyr)
library(tidyr)
#sum columns a-f
table1 <- df %>%
group_by(Person) %>%
summarise(across(a:f, sum, na.rm = TRUE))
#Remove duplicated values and get the data in separate columns
#for Rate and Street columns.
table2 <- df %>%
group_by(Person) %>%
mutate(across(c(Rate, Street), ~replace(., duplicated(.), NA))) %>%
select(Person, Rate, Street) %>%
filter(if_any(c(Rate, Street), ~!is.na(.))) %>%
mutate(col = row_number()) %>%
ungroup %>%
pivot_wider(names_from = col, values_from = c(Rate, Street)) %>%
select(where(~any(!is.na(.))))
#Join the two data to get final result
inner_join(table1, table2, by = 'Person')
# Person a b c d e f Rate_1 Rate_2 Street_1
# <chr> <int> <int> <int> <int> <int> <int> <int> <int> <chr>
#1 A 1 2 3 4 5 6 2 3 XYZ
#2 B 0 2 3 0 5 6 2 NA DEF
#3 C 1 2 3 4 5 6 1 NA DEF
data
It is helpful and easier to help when you share data in a reproducible format which can be copied directly. I have used the below data for the answer.
df <- structure(list(Person = c("A", "A", "A", "B", "B", "C"), Rate = c(2L,
2L, 3L, 2L, 2L, 1L), Street = c("XYZ", "XYZ", "XYZ", "DEF", "DEF",
"DEF"), a = c(1L, NA, NA, NA, NA, 1L), b = c(NA, 2L, NA, NA,
2L, 2L), c = c(3L, NA, NA, NA, 3L, 3L), d = c(4L, NA, NA, NA,
NA, 4L), e = c(5L, NA, NA, 5L, NA, 5L), f = c(NA, NA, 6L, NA,
6L, 6L)), row.names = c(NA, -6L), class = "data.frame")
I have this data frame:
df <- data.frame(
id = rep(1:4, each = 4),
status = c(
NA, "a", "c", "a",
NA, "b", "c", "c",
NA, NA, "a", "c",
NA, NA, "b", "b"),
stringsAsFactors = FALSE)
For each group (id), I aim to remove the rows with one or multiple leading NA in front of an "a" (in the column "status") but not in front of a "b".
The final data frame should look like this:
structure(list(
id = c(1L, 1L, 1L, 2L, 2L, 2L, 2L, 3L, 3L, 4L, 4L, 4L, 4L),
status = c("a", "c", "a", NA, "b", "c", "c", "a", "c", NA, NA, "b", "b")),
.Names = c("id", "status"), row.names = c(NA, -13L), class = "data.frame")
How do I do that?
Edit: alternatively, how would I do it to preserve other variables in the data frame such as the variable otherVar in the following example:
df2 <- data.frame(
id = rep(1:4, each = 4),
status = c(
NA, "a", "c", "a",
NA, "b", "c", "c",
NA, NA, "a", "c",
NA, NA, "b", "b"),
otherVar = letters[1:16],
stringsAsFactors = FALSE)
We can group by 'id', summarise the 'status' by pasteing the elements together, then use gsub to remove the NA before the 'a' and convert it to 'long' format with separate_rows
library(dplyr)
library(tidyr)
df %>%
group_by(id) %>%
summarise(status = gsub("(NA, ){1,}(?=a)", "", toString(status),
perl = TRUE)) %>%
separate_rows(status, convert = TRUE)
# A tibble: 13 x 2
# id status
# <int> <chr>
# 1 1 a
# 2 1 c
# 3 1 a
# 4 2 NA
# 5 2 b
# 6 2 c
# 7 2 c
# 8 3 a
# 9 3 c
#10 4 NA
#11 4 NA
#12 4 b
#13 4 b
Or using data.table with the same methodology
library(data.table)
out1 <- setDT(df)[, strsplit(gsub("(NA, ){1,}(?=a)", "",
toString(status), perl = TRUE), ", "), id]
setnames(out1, 'V1', "status")[]
# id status
# 1: 1 a
# 2: 1 c
# 3: 1 a
# 4: 2 NA
# 5: 2 b
# 6: 2 c
# 7: 2 c
# 8: 3 a
# 9: 3 c
#10: 4 NA
#11: 4 NA
#12: 4 b
#13: 4 b
Update
For the updated dataset 'df2'
i1 <- setDT(df2)[, .I[seq(which(c(diff((status %in% "a") +
rleid(is.na(status))) > 1), FALSE))] , id]$V1
df2[-i1]
# id status otherVar
# 1: 1 a b
# 2: 1 c c
# 3: 1 a d
# 4: 2 NA e
# 5: 2 b f
# 6: 2 c g
# 7: 2 c h
# 8: 3 a k
# 9: 3 c l
#10: 4 NA m
#11: 4 NA n
#12: 4 b o
#13: 4 b p
From zoo with na.locf and is.na, notice it assuming you data is ordered.
df[!(na.locf(df$status,fromLast = T)=='a'&is.na(df$status)),]
id status
2 1 a
3 1 c
4 1 a
5 2 <NA>
6 2 b
7 2 c
8 2 c
11 3 a
12 3 c
13 4 <NA>
14 4 <NA>
15 4 b
16 4 b
Here's a dplyr solution and a not as pretty base translation :
dplyr
library(dplyr)
df %>% group_by(id) %>%
filter(status[!is.na(status)][1]!="a" | !is.na(status))
# # A tibble: 13 x 2
# # Groups: id [4]
# id status
# <int> <chr>
# 1 1 a
# 2 1 c
# 3 1 a
# 4 2 <NA>
# 5 2 b
# 6 2 c
# 7 2 c
# 8 3 a
# 9 3 c
# 10 4 <NA>
# 11 4 <NA>
# 12 4 b
# 13 4 b
base
do.call(rbind,
lapply(split(df,df$id),
function(x) x[x$status[!is.na(x$status)][1]!="a" | !is.na(x$status),]))
# id status
# 1.2 1 a
# 1.3 1 c
# 1.4 1 a
# 2.5 2 <NA>
# 2.6 2 b
# 2.7 2 c
# 2.8 2 c
# 3.11 3 a
# 3.12 3 c
# 4.13 4 <NA>
# 4.14 4 <NA>
# 4.15 4 b
# 4.16 4 b
note
Will fail if not all NAs are leading because will remove all NAs from groups starting with "a" as a first non NA value.
I have a data frame that looks like this:
cat df1 df2 df3
1 1 NA 1 NA
2 1 NA 2 NA
3 1 NA 3 NA
4 2 1 NA NA
5 2 2 NA NA
6 2 3 NA NA
I want to populate df3 so that when cat = 1, df3 = df2 and when cat = 2, df3 = df1. However I am getting a few different error messages.
My current code looks like this:
df$df3[df$cat == 1] <- df$df2
df$df3[df$cat == 2] <- df$df1
Try this code:
df[df$cat==1,"df3"]<-df[df$cat==1,"df2"]
df[df$cat==2,"df3"]<-df[df$cat==1,"df1"]
The output:
df
cat df1 df2 df3
1 1 1 1 1
2 2 1 2 1
3 3 1 3 NA
4 4 2 NA NA
5 5 2 NA NA
6 5 2 NA NA
You can try
ifelse(df$cat == 1, df$df2, df$df1)
[1] 1 2 3 1 2 3
# saving
df$df3 <- ifelse(df$cat == 1, df$df2, df$df1)
# if there are other values than 1 and 2 you can try a nested ifelse
# that is setting other values to NA
ifelse(df$cat == 1, df$df2, ifelse(df$cat == 2, df$df1, NA))
# or you can try a tidyverse solution.
library(tidyverse)
df %>%
mutate(df3=case_when(cat == 1 ~ df2,
cat == 2 ~ df1))
cat df1 df2 df3
1 1 NA 1 1
2 1 NA 2 2
3 1 NA 3 3
4 2 1 NA 1
5 2 2 NA 2
6 2 3 NA 3
# data
df <- structure(list(cat = c(1L, 1L, 1L, 2L, 2L, 2L), df1 = c(NA, NA,
NA, 1L, 2L, 3L), df2 = c(1L, 2L, 3L, NA, NA, NA), df3 = c(NA,
NA, NA, NA, NA, NA)), .Names = c("cat", "df1", "df2", "df3"), class = "data.frame", row.names = c("1",
"2", "3", "4", "5", "6"))
I am having issue with rearranging some data.
The original data is:
structure(list(id = 1:3, artery.1 = structure(c(1L, 1L, 2L), .Label = c("a",
"b"), class = "factor"), artery.2 = structure(c(1L, NA, 2L), .Label = c("b",
"c"), class = "factor"), artery.3 = structure(c(1L, NA, 2L), .Label = c("c",
"d"), class = "factor"), artery.4 = structure(c(NA, NA, 1L), .Label = "e", class = "factor"), artery.5 = structure(c(NA, NA, 1L), .Label = "f", class = "factor"),
diameter.1 = c(3L, 2L, 1L), diameter.2 = c(2L, NA, 2L), diameter.3 = c(3L,
NA, 3L), diameter.4 = c(NA, NA, 4L), diameter.5 = c(NA, NA,
5L)), .Names = c("id", "artery.1", "artery.2", "artery.3",
"artery.4", "artery.5", "diameter.1", "diameter.2", "diameter.3",
"diameter.4", "diameter.5"), class = "data.frame", row.names = c(NA,
-3L))
# id artery.1 artery.2 artery.3 artery.4 artery.5 diameter.1 diameter.2 diameter.3 diameter.4 diameter.5
# 1 1 a b c <NA> <NA> 3 2 3 NA NA
# 2 2 a <NA> <NA> <NA> <NA> 2 NA NA NA NA
# 3 3 b c d e f 1 2 3 4 5
I would like to get to this:
structure(list(id = 1:3, a = c(3L, 2L, NA), b = c(2L, NA, 1L),
c = c(3L, NA, 2L), d = c(NA, NA, 3L), e = c(NA, NA, 4L),
f = c(NA, NA, 5L)), .Names = c("id", "a", "b", "c", "d",
"e", "f"), class = "data.frame", row.names = c(NA, -3L))
# id a b c d e f
# 1 1 3 2 3 NA NA NA
# 2 2 2 NA NA NA NA NA
# 3 3 NA 1 2 3 4 5
Basically, a to f represents arteries and the numerical values represent the corresponding diameter. Each row represents a patient.
Is there a neat way to sort this dataframe out?
Modern tidyr makes the solution even more succinct via the pivot_ functions:
library(dplyr)
library(tidyr)
df %>%
pivot_longer(-id, names_pattern = '(artery|diameter)\\.(\\d+)', names_to = c('.value', NA)) %>%
filter(!is.na(artery)) %>%
pivot_wider(names_from = artery, values_from = diameter)
id a b c d e f
<int> <int> <int> <int> <int> <int> <int>
1 1 3 2 3 NA NA NA
2 2 2 NA NA NA NA NA
3 3 NA 1 2 3 4 5
Here is the older solution, which uses the deprecated gather and spread functions:
library(dplyr)
library(tidyr)
new.df <- gather(df, variable, value, artery.1:diameter.5) %>%
separate(variable, c('variable', 'num')) %>%
spread(variable, value) %>%
subset(!is.na(artery)) %>%
mutate(diameter = as.numeric(diameter)) %>%
select(-num) %>%
spread(artery, diameter)
Output:
id a b c d e f
1 1 3 2 3 NA NA NA
2 2 2 NA NA NA NA NA
3 3 NA 1 2 3 4 5
Or using melt/dcast combination with data.table while selecting variables using regex in the patterns function
library(data.table) #v>=1.9.6
dcast(melt(setDT(df),
id = "id",
measure = patterns("artery", "diameter")),
id ~ value1,
sum,
value.var = "value2",
subset = .(!is.na(value2)),
fill = NA)
# id a b c d e f
# 1: 1 3 2 3 NA NA NA
# 2: 2 2 NA NA NA NA NA
# 3: 3 NA 1 2 3 4 5
As you can see, both melt and dcast are very flexible and you can use regex, specify a subset, pass multiple functions and specify how you want to fill missing values.
You can use xtabs with reshape from base R. Use the latter to transform data to long format and use the former to get the count table:
xtabs(diameter ~ id + artery, reshape(df, varying = 2:11, sep = '.', dir = "long"))
# artery
#id a b c d e f
# 1 3 2 3 0 0 0
# 2 2 0 0 0 0 0
# 3 0 1 2 3 4 5
This can be done with two reshape() calls. First, we can longify both artery and diameter on id, then widen with artery as the time variable. To prevent a column of NAs, we also must subset out rows with NA values for artery in the intermediate frame.
reshape(subset(reshape(df,dir='l',varying=setdiff(names(df),'id'),timevar=NULL),!is.na(artery)),dir='w',timevar='artery');
## id diameter.a diameter.b diameter.c diameter.d diameter.e diameter.f
## 1.1 1 3 2 3 NA NA NA
## 2.1 2 2 NA NA NA NA NA
## 3.1 3 NA 1 2 3 4 5
The diameter. prefixes can be removed afterward, if desired. However, an advantage of this solution is that it would be capable of preserving multiple column sets, whereas the xtabs() solution cannot. The prefixes would be essential to distinguish the column sets in that case.