Related
i have a data.frame with columns and rows. how could i replace NA values so that it would be the average of the first value before and after that cell in that column?
for example:
1. 1 2 3
2. 4 NA 7
3. 9 NA 8
4. 1 5 6
I need the first NA to be - (5+2)/2=3.5
and the second to be (3.5+5)/2=4.25
Lets create some sample data and transform it to data.table:
require(data.table)
require(zoo)
dat <- data.frame(a = c(1, 2, NA, 4))
setDT(dat)
Now, using the zoo::na.approx function we can impute the missing values.
dat[, newA:= na.approx(a, rule = 2)]
Output:
a newA
1: 1 1
2: 2 2
3: NA 3
4: 4 4
I want to use data.table to incrementally find out new elements i.e. for every row, I'd see whether values in list have been seen before. If they are, we will ignore them. If not, we will select them.
I was able to wrap elements by group in a list, but I am unsure how I can find incremental differences.
Here's my attempt:
df = data.table::data.table(id = c('A','B','C','A','B','A','A','A','D','E','E','E'),
Value = c(1,2,3,4,3,5,2,3,7,2,3,9))
df_wrapped=df[,.(Values=(list(unique(Value)))), by=id]
expected_output = data.table::data.table(id = c("A","B","C","D","E"),
Value = list(c(1,4,5,2,3),c(2,3),c(3),c(7),c(2,3,9)),
Diff=list(c(1,4,5,2,3),c(NA),c(NA),c(7),c(9)),
Count = c(5,0,0,1,1))
Thoughts about expected output:
For the first row, all elements are unique. So, we will include them in Diff column.
In the second row, 2,3 have occurred in row 1. So, we will ignore them. Ditto for row 3.
Similarly, 7 and 9 are seen for the first time in row 4 and 5, so we will include them.
Here's visual representation:
expected_output
id Value Diff Count
A 1,4,5,2,3 1,4,5,2,3 5
B 2,3 NA 0
C 3 NA 0
D 7 7 1
E 2,3,9 9 1
I'd appreciate any thoughts. I am only looking for data.table based solutions because of performance issues in my original dataset.
I am not sure why you specifically need to put them in a list, but otherwise I wrote a small piece that could help you.
df = data.table::data.table(id = c('A','B','C','A','B','A','A','A','D','E','E','E'),
Value = c(1,2,3,4,3,5,2,3,7,2,3,9))
df = df[order(id, Value)]
df = df[duplicated(Value) == FALSE, diff := Value][]
df = df[, count := uniqueN(diff, na.rm = TRUE), by = id]
The outcome would be:
> df
id Value diff count
1: A 1 1 5
2: A 2 2 5
3: A 3 3 5
4: A 4 4 5
5: A 5 5 5
6: B 2 NA 0
7: B 3 NA 0
8: C 3 NA 0
9: D 7 7 1
10: E 2 NA 1
11: E 3 NA 1
12: E 9 9 1
Hope this helps, or at least get you started.
Here is another possible approach:
library(data.table)
df = data.table(
id = c('A','B','C','A','B','A','A','A','D','E','E','E'),
Value = c(1,2,3,4,3,5,2,3,7,2,3,9))
valset <- c()
df[, {
d <- setdiff(Value, valset)
valset <- unique(c(valset, Value))
.(Values=.(Value), Diff=.(d), Count=length(d))
},
by=.(id)]
output:
id Values Diff Count
1: A 1,4,5,2,3 1,4,5,2,3 5
2: B 2,3 0
3: C 3 0
4: D 7 7 1
5: E 2,3,9 9 1
I have a large data set - around 32mil rows. I have information on the telephone number, the origin of the call, and the destination.
For each telephone number, I want to count the number of times it appeared either as Origin or as Destination.
An example data table is as follows:
library(data.table)
dt <- data.table(Tel=seq(1,5,1), Origin=seq(1,5,1), Destination=seq(3,7,1))
Tel Origin Destination
1: 1 1 3
2: 2 2 4
3: 3 3 5
4: 4 4 6
5: 5 5 7
I have working code, but it takes too long for my data since it involves a for loop. How can I optimize it?
Here it is:
for (i in unique(dt$Tel)){
index <- (dt$Origin == i | dt$Destination == i)
dt[dt$Tel ==i, "N"] <- sum(index)
}
Result:
Tel Origin Destination N
1: 1 1 3 1
2: 2 2 4 1
3: 3 3 5 2
4: 4 4 6 2
5: 5 5 7 2
Where N tells that Tel=1 appears 1, Tel=2 appears 1, Tel=3,4 and 5 each appear 2 times.
We can do a melt and match
dt[, N := melt(dt, id.var = "Tel")[, tabulate(match(value, Tel))]]
Or another option is to loop through the columns 2 and 3, use %in% to check whether the values in 'Tel' are present, then with Reduce and + get the sum of logical elements for each 'Tel', assign (:=) the values to 'N'
dt[, N := Reduce(`+`, lapply(.SD, function(x) Tel %in% x)), .SDcols = 2:3]
dt
# Tel Origin Destination N
#1: 1 1 3 1
#2: 2 2 4 1
#3: 3 3 5 2
#4: 4 4 6 2
#5: 5 5 7 2
A second method constructs a temporary data.table which is then joins to the original. This is longer and likely less efficient than #akrun's, but can be useful to see.
# get temporary data.table as the sum of origin and destination frequencies
temp <- setnames(data.table(table(unlist(dt[, .(Origin, Destination)], use.names=FALSE))),
c("Tel", "N"))
# turn the variables into integers (Tel is the name of the table above, and thus character)
temp <- temp[, lapply(temp, as.integer)]
Now, join the original table on
dt <- temp[dt, on="Tel"]
dt
Tel N Origin Destination
1: 1 1 1 3
2: 2 1 2 4
3: 3 2 3 5
4: 4 2 4 6
5: 5 2 5 7
You can get the desired column order using setcolorder
setcolorder(dt, c("Tel", "Origin", "Destination", "N"))
I want to reshape a data.table, and include the historic (cumulative summed) information for each variable. The No variable indicates the chronological order of measurements for object ID. At each measurement additional information is found. I want to aggregate the known information at each timestamp No for object ID.
Let me demonstrate with an example:
For the following data.table:
df <- data.table(ID=c(1,1,1,2,2,2,2),
No=c(1,2,3,1,2,3,4),
Variable=c('a','b', 'a', 'c', 'a', 'a', 'b'),
Value=c(2,1,3,3,2,1,5))
df
ID No Variable Value
1: 1 1 a 2
2: 1 2 b 1
3: 1 3 a 3
4: 2 1 c 3
5: 2 2 a 2
6: 2 3 a 1
7: 2 4 b 5
I want to reshape it to this:
ID No a b c
1: 1 1 2 NA NA
2: 1 2 2 1 NA
3: 1 3 5 1 NA
4: 2 1 NA NA 3
5: 2 2 2 NA 3
6: 2 3 3 NA 3
7: 2 4 3 5 3
So the summed values of Value, per Variable by (ID, No), cumulative over No.
I can get the result without the cumulative part by doing
dcast(df, ID+No~Variable, value.var="Value")
which results in the non-cumulative variant:
ID No a b c
1: 1 1 2 NA NA
2: 1 2 NA 1 NA
3: 1 3 3 NA NA
4: 2 1 NA NA 3
5: 2 2 2 NA NA
6: 2 3 1 NA NA
7: 2 4 NA 5 NA
Any ideas how to make this cumulative? The original data.table has over 250,000 rows, so efficiency matters.
EDIT: I just used a,b,c as an example, the original file has about 40 different levels. Furthermore, the NAs are important; there are also Value-values of 0, which means something else than NA
POSSIBLE SOLUTION
Okay, so I've found a working solution. It is far from efficient, since it enlarges the original table.
The idea is to duplicate each row TotalNo - No times, where TotalNo is the maximum No per ID. Then the original dcast function can be used to extract the dataframe. So in code:
df[,TotalNo := .N, by=ID]
df2 <- df[rep(seq(nrow(df)), (df$TotalNo - df$No + 1))] #create duplicates
df3 <- df2[order(ID, No)]#, No:= seq_len(.N), by=.(ID, No)]
df3[,No:= seq(from=No[1], to=TotalNo[1], by=1), by=.(ID, No)]
df4<- dcast(df3,
formula = ID + No ~ Variable,
value.var = "Value", fill=NA, fun.aggregate = sum)
It is not really nice, because the creation of duplicates uses more memory. I think it can be further optimized, but so far it works for my purposes. In the sample code it goes from 7 rows to 16 rows, in the original file from 241,670 rows to a whopping 978,331. That's over a factor 4 larger.
SOLUTION
Eddi has improved my solution in computing time in the full dataset (2.08 seconds of Eddi versus 4.36 seconds of mine). Those are numbers I can work with! Thanks everybody!
Your solution is good, but you're adding too many rows, that are unnecessary if you compute the cumsum beforehand:
# add useful columns
df[, TotalNo := .N, by = ID][, CumValue := cumsum(Value), by = .(ID, Variable)]
# do a rolling join to extend the missing values, and then dcast
dcast(df[df[, .(No = seq(No[1], TotalNo[1])), by = .(ID, Variable)],
on = c('ID', 'Variable', 'No'), roll = TRUE],
ID + No ~ Variable, value.var = 'CumValue')
# ID No a b c
#1: 1 1 2 NA NA
#2: 1 2 2 1 NA
#3: 1 3 5 1 NA
#4: 2 1 NA NA 3
#5: 2 2 2 NA 3
#6: 2 3 3 NA 3
#7: 2 4 3 5 3
Here's a standard way:
library(zoo)
df[, cv := cumsum(Value), by = .(ID, Variable)]
DT = dcast(df, ID + No ~ Variable, value.var="cv")
lvls = sort(unique(df$Variable))
DT[, (lvls) := lapply(.SD, na.locf, na.rm = FALSE), by=ID, .SDcols=lvls]
ID No a b c
1: 1 1 2 NA NA
2: 1 2 2 1 NA
3: 1 3 5 1 NA
4: 2 1 NA NA 3
5: 2 2 2 NA 3
6: 2 3 3 NA 3
7: 2 4 3 5 3
One alternative way to do it is using a custom built cumulative sum function. This is exactly the method in #David Arenburg's comment, but substitutes in a custom cumulative summary function.
EDIT: Using #eddi's much more efficient custom cumulative sum function.
cumsum.na <- function(z){
Reduce(function(x, y) if (is.na(x) && is.na(y)) NA else sum(x, y, na.rm = T), z, accumulate = T)
}
cols <- sort(unique(df$Variable))
res <- dcast(df, ID + No ~ Variable, value.var = "Value")[, (cols) := lapply(.SD, cumsum.na), .SDcols = cols, by = ID]
res
ID No a b c
1: 1 1 2 NA NA
2: 1 2 2 1 NA
3: 1 3 5 1 NA
4: 2 1 NA NA 3
5: 2 2 2 NA 3
6: 2 3 3 NA 3
7: 2 4 3 5 3
This definitely isn't the most efficient, but it gets the job done and gives you an admittedly very slow very slow cumulative summary function that handles NAs the way you want to.
I have monthly data in one data.table and annual data in another data.table and now I want to match the annual data to the respective observation in the monthly data.
My approach is as follows: Duplicating the annual data for every month and then join the monthly and annual data. And now I have a question regarding the duplication of rows. I know how to do it, but I'm not sure if it is the best way to do it, so some opinions would be great.
Here is an exemplatory data.table DT for my annual data and how I currently duplicate:
library(data.table)
DT <- data.table(ID = paste(rep(c("a", "b"), each=3), c(1:3, 1:3), sep="_"),
values = 10:15,
startMonth = seq(from=1, by=2, length=6),
endMonth = seq(from=3, by=3, length=6))
DT
ID values startMonth endMonth
[1,] a_1 10 1 3
[2,] a_2 11 3 6
[3,] a_3 12 5 9
[4,] b_1 13 7 12
[5,] b_2 14 9 15
[6,] b_3 15 11 18
#1. Alternative
DT1 <- DT[, list(MONTH=startMonth:endMonth), by="ID"]
setkey(DT, ID)
setkey(DT1, ID)
DT1[DT]
ID MONTH values startMonth endMonth
a_1 1 10 1 3
a_1 2 10 1 3
a_1 3 10 1 3
a_2 3 11 3 6
[...]
The last join is exactly what I want. However, DT[, list(MONTH=startMonth:endMonth), by="ID"] already does everything I want except adding the other columns to DT, so I was wondering if I could get rid of the last three rows in my code, i.e. the setkey and join operations. It turns out, you can, just do the following:
#2. Alternative: More intuitiv and just one line of code
DT[, list(MONTH=startMonth:endMonth, values, startMonth, endMonth), by="ID"]
ID MONTH values startMonth endMonth
a_1 1 10 1 3
a_1 2 10 1 3
a_1 3 10 1 3
a_2 3 11 3 6
...
This, however, only works because I hardcoded the column names into the list expression. In my real data, I do not know the names of all columns in advance, so I was wondering if I could just tell data.table to return the column MONTH that I compute as shown above and all the other columns of DT. .SD seemed to be able to do the trick, but:
DT[, list(MONTH=startMonth:endMonth, .SD), by="ID"]
Error in `[.data.table`(DT, , list(YEAR = startMonth:endMonth, .SD), by = "ID") :
maxn (4) is not exact multiple of this j column's length (3)
So to summarize, I know how it's been done, but I was just wondering if this is the best way to do it because I'm still struggling a little bit with the syntax of data.table and often read in posts and on the wiki that there are good and bads ways of doing things. Also, I don't quite get why I get an error when using .SD. I thought it is just any easy way to tell data.table that you want all columns. What do I miss?
Looking at this I realized that the answer was only possible because ID was a unique key (without duplicates). Here is another answer with duplicates. But, by the way, some NA seem to creep in. Could this be a bug? I'm using v1.8.7 (commit 796).
library(data.table)
DT <- data.table(x=c(1,1,1,1,2,2,3),y=c(1,1,2,3,1,1,2))
DT[,rep:=1L][c(2,7),rep:=c(2L,3L)] # duplicate row 2 and triple row 7
DT[,num:=1:.N] # to group each row by itself
DT
x y rep num
1: 1 1 1 1
2: 1 1 2 2
3: 1 2 1 3
4: 1 3 1 4
5: 2 1 1 5
6: 2 1 1 6
7: 3 2 3 7
DT[,cbind(.SD,dup=1:rep),by="num"]
num x y rep dup
1: 1 1 1 1 1
2: 2 1 1 1 NA # why these NA?
3: 2 1 1 2 NA
4: 3 1 2 1 1
5: 4 1 3 1 1
6: 5 2 1 1 1
7: 6 2 1 1 1
8: 7 3 2 3 1
9: 7 3 2 3 2
10: 7 3 2 3 3
Just for completeness, a faster way is to rep the row numbers and then take the subset in one step (no grouping and no use of cbind or .SD) :
DT[rep(num,rep)]
x y rep num
1: 1 1 1 1
2: 1 1 2 2
3: 1 1 2 2
4: 1 2 1 3
5: 1 3 1 4
6: 2 1 1 5
7: 2 1 1 6
8: 3 2 3 7
9: 3 2 3 7
10: 3 2 3 7
where in this example data the column rep happens to be the same name as the rep() base function.
Great question. What you tried was very reasonable. Assuming you're using v1.7.1 it's now easier to make list columns. In this case it's trying to make one list column out of .SD (3 items) alongside the MONTH column of the 2nd group (4 items). I'll raise it as a bug [EDIT: now fixed in v1.7.5], thanks.
In the meantime, try :
DT[, cbind(MONTH=startMonth:endMonth, .SD), by="ID"]
ID MONTH values startMonth endMonth
a_1 1 10 1 3
a_1 2 10 1 3
a_1 3 10 1 3
a_2 3 11 3 6
...
Also, just to check you've seen roll=TRUE? Typically you'd have just one startMonth column (irregular with gaps) and then just roll join to it. Your example data has overlapping month ranges though, so that complicates it.
Here is a function I wrote which mimics disaggregate (I needed something that handled complex data). It might be useful for you, if it isn't overkill. To expand only rows, set the argument fact to c(1,12) where 12 would be for 12 'month' rows for each 'year' row.
zexpand<-function(inarray, fact=2, interp=FALSE, ...) {
fact<-as.integer(round(fact))
switch(as.character(length(fact)),
'1' = xfact<-yfact<-fact,
'2'= {xfact<-fact[1]; yfact<-fact[2]},
{xfact<-fact[1]; yfact<-fact[2];warning(' fact is too long. First two values used.')})
if (xfact < 1) { stop('fact[1] must be > 0') }
if (yfact < 1) { stop('fact[2] must be > 0') }
# new nonloop method, seems to work just ducky
bigtmp <- matrix(rep(t(inarray), each=xfact), nrow(inarray), ncol(inarray)*xfact, byr=T)
#does column expansion
bigx <- t(matrix(rep((bigtmp),each=yfact),ncol(bigtmp),nrow(bigtmp)*yfact,byr=T))
return(invisible(bigx))
}
The fastest and most succinct way of doing it:
DT[rep(1:nrow(DT), endMonth - startMonth)]
We can also enumerate by group by:
dd <- DT[rep(1:nrow(DT), endMonth - startMonth)]
dd[, nn := 1:.N, by = ID]
dd