How can I use apply or a related function to create a new data frame that contains the results of the row averages of each pair of columns in a very large data frame?
I have an instrument that outputs n replicate measurements on a large number of samples, where each single measurement is a vector (all measurements are the same length vectors). I'd like to calculate the average (and other stats) on all replicate measurements of each sample. This means I need to group n consecutive columns together and do row-wise calculations.
For a simple example, with three replicate measurements on two samples, how can I end up with a data frame that has two columns (one per sample), one that is the average each row of the replicates in dat$a, dat$b and dat$c and one that is the average of each row for dat$d, dat$e and dat$f.
Here's some example data
dat <- data.frame( a = rnorm(16), b = rnorm(16), c = rnorm(16), d = rnorm(16), e = rnorm(16), f = rnorm(16))
a b c d e f
1 -0.9089594 -0.8144765 0.872691548 0.4051094 -0.09705234 -1.5100709
2 0.7993102 0.3243804 0.394560355 0.6646588 0.91033497 2.2504104
3 0.2963102 -0.2911078 -0.243723116 1.0661698 -0.89747522 -0.8455833
4 -0.4311512 -0.5997466 -0.545381175 0.3495578 0.38359390 0.4999425
5 -0.4955802 1.8949285 -0.266580411 1.2773987 -0.79373386 -1.8664651
6 1.0957793 -0.3326867 -1.116623982 -0.8584253 0.83704172 1.8368212
7 -0.2529444 0.5792413 -0.001950741 0.2661068 1.17515099 0.4875377
8 1.2560402 0.1354533 1.440160168 -2.1295397 2.05025701 1.0377283
9 0.8123061 0.4453768 1.598246016 0.7146553 -1.09476532 0.0600665
10 0.1084029 -0.4934862 -0.584671816 -0.8096653 1.54466019 -1.8117459
11 -0.8152812 0.9494620 0.100909570 1.5944528 1.56724269 0.6839954
12 0.3130357 2.6245864 1.750448404 -0.7494403 1.06055267 1.0358267
13 1.1976817 -1.2110708 0.719397607 -0.2690107 0.83364274 -0.6895936
14 -2.1860098 -0.8488031 -0.302743475 -0.7348443 0.34302096 -0.8024803
15 0.2361756 0.6773727 1.279737692 0.8742478 -0.03064782 -0.4874172
16 -1.5634527 -0.8276335 0.753090683 2.0394865 0.79006103 0.5704210
I'm after something like this
X1 X2
1 -0.28358147 -0.40067128
2 0.50608365 1.27513471
3 -0.07950691 -0.22562957
4 -0.52542633 0.41103139
5 0.37758930 -0.46093340
6 -0.11784382 0.60514586
7 0.10811540 0.64293184
8 0.94388455 0.31948189
9 0.95197629 -0.10668118
10 -0.32325169 -0.35891702
11 0.07836345 1.28189698
12 1.56269017 0.44897971
13 0.23533617 -0.04165384
14 -1.11251880 -0.39810121
15 0.73109533 0.11872758
16 -0.54599850 1.13332286
which I did with this, but is obviously no good for my much larger data frame...
data.frame(cbind(
apply(cbind(dat$a, dat$b, dat$c), 1, mean),
apply(cbind(dat$d, dat$e, dat$f), 1, mean)
))
I've tried apply and loops and can't quite get it together. My actual data has some hundreds of columns.
This may be more generalizable to your situation in that you pass a list of indices. If speed is an issue (large data frame) I'd opt for lapply with do.call rather than sapply:
x <- list(1:3, 4:6)
do.call(cbind, lapply(x, function(i) rowMeans(dat[, i])))
Works if you just have col names too:
x <- list(c('a','b','c'), c('d', 'e', 'f'))
do.call(cbind, lapply(x, function(i) rowMeans(dat[, i])))
EDIT
Just happened to think maybe you want to automate this to do every three columns. I know there's a better way but here it is on a 100 column data set:
dat <- data.frame(matrix(rnorm(16*100), ncol=100))
n <- 1:ncol(dat)
ind <- matrix(c(n, rep(NA, 3 - ncol(dat)%%3)), byrow=TRUE, ncol=3)
ind <- data.frame(t(na.omit(ind)))
do.call(cbind, lapply(ind, function(i) rowMeans(dat[, i])))
EDIT 2
Still not happy with the indexing. I think there's a better/faster way to pass the indexes. here's a second though not satisfying method:
n <- 1:ncol(dat)
ind <- data.frame(matrix(c(n, rep(NA, 3 - ncol(dat)%%3)), byrow=F, nrow=3))
nonna <- sapply(ind, function(x) all(!is.na(x)))
ind <- ind[, nonna]
do.call(cbind, lapply(ind, function(i)rowMeans(dat[, i])))
A similar question was asked here by #david: averaging every 16 columns in r (now closed), which I answered by adapting #TylerRinker's answer above, following a suggestion by #joran and #Ben. Because the resulting function might be of help to OP or future readers, I am copying that function here, along with an example for OP's data.
# Function to apply 'fun' to object 'x' over every 'by' columns
# Alternatively, 'by' may be a vector of groups
byapply <- function(x, by, fun, ...)
{
# Create index list
if (length(by) == 1)
{
nc <- ncol(x)
split.index <- rep(1:ceiling(nc / by), each = by, length.out = nc)
} else # 'by' is a vector of groups
{
nc <- length(by)
split.index <- by
}
index.list <- split(seq(from = 1, to = nc), split.index)
# Pass index list to fun using sapply() and return object
sapply(index.list, function(i)
{
do.call(fun, list(x[, i], ...))
})
}
Then, to find the mean of the replicates:
byapply(dat, 3, rowMeans)
Or, perhaps the standard deviation of the replicates:
byapply(dat, 3, apply, 1, sd)
Update
by can also be specified as a vector of groups:
byapply(dat, c(1,1,1,2,2,2), rowMeans)
mean for rows from vectors a,b,c
rowMeans(dat[1:3])
means for rows from vectors d,e,f
rowMeans(dat[4:6])
all in one call you get
results<-cbind(rowMeans(dat[1:3]),rowMeans(dat[4:6]))
if you only know the names of the columns and not the order then you can use:
rowMeans(cbind(dat["a"],dat["b"],dat["c"]))
rowMeans(cbind(dat["d"],dat["e"],dat["f"]))
#I dont know how much damage this does to speed but should still be quick
The rowMeans solution will be faster, but for completeness here's how you might do this with apply:
t(apply(dat,1,function(x){ c(mean(x[1:3]),mean(x[4:6])) }))
Inspired by #joran's suggestion I came up with this (actually a bit different from what he suggested, though the transposing suggestion was especially useful):
Make a data frame of example data with p cols to simulate a realistic data set (following #TylerRinker's answer above and unlike my poor example in the question)
p <- 99 # how many columns?
dat <- data.frame(matrix(rnorm(4*p), ncol = p))
Rename the columns in this data frame to create groups of n consecutive columns, so that if I'm interested in the groups of three columns I get column names like 1,1,1,2,2,2,3,3,3, etc or if I wanted groups of four columns it would be 1,1,1,1,2,2,2,2,3,3,3,3, etc. I'm going with three for now (I guess this is a kind of indexing for people like me who don't know much about indexing)
n <- 3 # how many consecutive columns in the groups of interest?
names(dat) <- rep(seq(1:(ncol(dat)/n)), each = n, len = (ncol(dat)))
Now use apply and tapply to get row means for each of the groups
dat.avs <- data.frame(t(apply(dat, 1, tapply, names(dat), mean)))
The main downsides are that the column names in the original data are replaced (though this could be overcome by putting the grouping numbers in a new row rather than the colnames) and that the column names are returned by the apply-tapply function in an unhelpful order.
Further to #joran's suggestion, here's a data.table solution:
p <- 99 # how many columns?
dat <- data.frame(matrix(rnorm(4*p), ncol = p))
dat.t <- data.frame(t(dat))
n <- 3 # how many consecutive columns in the groups of interest?
dat.t$groups <- as.character(rep(seq(1:(ncol(dat)/n)), each = n, len = (ncol(dat))))
library(data.table)
DT <- data.table(dat.t)
setkey(DT, groups)
dat.av <- DT[, lapply(.SD,mean), by=groups]
Thanks everyone for your quick and patient efforts!
There is a beautifully simple solution if you are interested in applying a function to each unique combination of columns, in what known as combinatorics.
combinations <- combn(colnames(df),2,function(x) rowMeans(df[x]))
To calculate statistics for every unique combination of three columns, etc., just change the 2 to a 3. The operation is vectorized and thus faster than loops, such as the apply family functions used above. If the order of the columns matters, then you instead need a permutation algorithm designed to reproduce ordered sets: combinat::permn
Related
I have a group of integers, as in this R data.frame:
set.seed(1)
df <- data.frame(id = paste0("id",1:100), length = as.integer(runif(100,10000,1000000)), stringsAsFactors = F)
So each element has an id and a length.
I'd like to split df into two data.frames with approximately equal sums of length.
Any idea of an R function to achieve that?
I thought that Hmisc's cut2 might do it but I don't think that's its intended use:
library(Hmisc) # cut2
ll <- split(df, cut2(df$length, g=2))
> sum(ll[[1]]$length)
[1] 14702139
> sum(ll[[2]]$length)
[1] 37564671
It's called Bin pack problem. https://en.wikipedia.org/wiki/Bin_packing_problem this link may be helpful.
Using BBmisc::binPack function,
df$bins <- binPack(df$length, sum(df$length)/2 + 1)
tapply(df$length, df$bins, sum)
results like
1 2 3
25019106 24994566 26346
Now since you want two groups,
dummy$bins[dummy$bins == 3] <- 2 #because labeled as 2's sum is smaller
result is
1 2
25019106 25020912
I'm building a little function in R that takes size measurements from several species and several sites, combines all the data by site (lumping many species together), and then computes some statistics on those combined data.
Here is some simplistic sample data:
SiteID <- rep(c("D00002", "D00003", "D00004"), c(5, 2, 3))
SpeciesID <- c("CHIL", "CHIP", "GAM", "NZMS", "LUMB", "CHIL", "SIMA", "CHIP", "CHIL", "NZMS")
Counts <- data.frame(matrix(sample(0:99,200, replace = TRUE), nrow = 10, ncol = 20))
colnames(Counts) <- paste0('B', 1:20)
spec <- cbind(SiteID, SpeciesID, Counts)
stat1 <- data.frame(unique(SiteID))
colnames(stat1) <- 'SiteID'
stat1$Mean <- NA
Here is the function, which creates a list, lsize1, where each list element is a vector of the sizes (B1 to B20) for a given SpeciesID in a given SiteID, multiplied by the number of counts for each size class. From this, the function creates a list, lsize2, which combines list elements from lsize1 that have the same SiteID. Finally, it gets the mean of each element in lsize2 (i.e., the average size of an individual for each SiteID, regardless of SpeciesID), and outputs that as a result.
fsize <- function(){
specB <- spec[, 3:22]
lsize1 <- apply(specB, 1, function(x) rep(1:20, x))
names(lsize1) <- spec$SiteID
lsize2 <- sapply(unique(names(lsize1)), function(x) unlist(lsize1[names(lsize1) == x], use.names = FALSE), simplify = FALSE)
stat1[stat1$SiteID %in% names(lsize2), 'Mean'] <- round(sapply(lsize2, mean), 2)
return(stat1)
}
In creating this function, I followed the suggestion here: combine list elements based on element names, which gets at the crux of my problem: combining list elements based on some criteria in common (in my case, combining all elements from the same SiteID). The function works as intended, but my question is if there's a way to make it substantially faster?
Note: for my actual data set, which is ~40,000 rows in length, I find that the function runs in ~ 0.7 seconds, with the most time consuming step being the creation of lsize2 (~ 0.5 seconds). I need to run this function many, many times, with different permutations and subsets of the data, so I'm hoping there's a way to cut this processing time down significantly.
There shouldn't be any need for loops here. Here's one attempt:
tmp <- data.frame(spec["SiteID"], sums = rowSums(specB * col(specB)), counts=rowSums(specB) )
tmp <- aggregate(. ~ SiteID, tmp, sum)
tmp$avg <- tmp$sums / tmp$counts
tmp
# SiteID sums counts avg
#1 D00002 46254 4549 10.16795
#2 D00003 20327 1810 11.23039
#3 D00004 29651 2889 10.26341
Compare:
fsize()
# SiteID Mean
#1 D00002 10.17
#2 D00003 11.23
#3 D00004 10.26
This code essentially multiplies each value by it's index (col(specB)), then aggregates the sums and counts by SiteID. This logic should be relatively transferable to other methods (data.table/dplyr) as well. E.g.: in data.table:
setDT(spec)
spec[, .(avg = sum(.SD * col(.SD)) / sum(unlist(.SD))), by=SiteID, .SDcols=B1:B20]
# SiteID avg
#1: D00002 10.16795
#2: D00003 11.23039
#3: D00004 10.26341
See how addition works over components:
a<-1:3
a+a #Gives (1+1), (2+2), (3+3)
I've considered using loops over argument lengths or transforming them into a data.frame and then using apply but I have the intuition there is a more efficient way of going about this.
Specifically, I'd like to calculate the mean of each set of components ignoring zero values, like so:
function(x) {
mean(x[x!=0])
}
Except x would be the i-th components of an arbitrary amount of arguments.
If I understand correctly, mapply or its wrapper Map would work fairly well here.
mapply(function(...) {temp <- c(...); mean(temp[temp != 0])}, 1:10, 11:20)
[1] 6 7 8 9 10 11 12 13 14 15
With mapply, the given function is applied to the collection of the first elements of each vector, then the collection of the second elements and so on. The function creates a new vector with c and then calculates the mean for all non-zero elements. This function returns an atomic vector.
Map(function(...) {temp <- c(...); mean(temp[temp != 0])}, 1:10, 11:20)
returns a list instead. This could be wrapped in unlist to return a vector.
If we need to do this sequentially from multiple vectors
Reduce(`+`, listofvectors)
Or rbind or cbind it to create a matrix and then do the colSums or rowSums
colSums(m1)
Update
Regarding the second part of the question (not clear), if it is to get the mean of individual vectors in a list excluding the 0 value
sapply(listofvectors, function(x) mean(x[x!=0]))
Or if we need the mean of sequence of elements in the matrix (created by rbinding the vectors), then replace the 0 values with NA, and get the colMeans with na.rm = TRUE
colMeans(replace(m1, m1==0, NA), na.rm = TRUE)
colMeans(replace(m2, m2==0, NA), na.rm = TRUE)
#[1] 6 7 8 9 10 11 12 13 14 15
NOTE: The colMeans and matrix approach are vectorized. No looping done here
data
a1 <- 1:5
b1 <- 6:10
c1 <- 11:15
listofvectors <- list(a1, b1, c1)
m1 <- rbind(a1, b1, c1)
m2 <- rbind(1:10, 11:20)
I have managed to read in a data file, and subset out the 2 columns of info that I want to work with. I am now stuck because I need to split the data into chunks of varying sizes and apply a function (mean, sd) to them, save the chunks and plot the sd from each. Otherwise known generally as block averaging. Right now I have a data frame with 2 columns and 10005 rows. The head of it looks like this:
Frame CA
1 0.773
Is there an efficient way that I could subset pieces of the data from a:b so that I can dictate how the data is broken up by the "Frame" column? I have found really good answers on here but I am not sure what they mean fully or if they would work.
chunk <- function(x, n)
(mapply(function(a, b) (x[a:b]), seq.int(from=1, to=length(x), by=n),
pmin(seq.int(from=1, to=length(x), by=n)+(n-1),
length(x)), SIMPLIFY=FALSE))
I'm not sure if it is what you're looking for but with closure, a data frame can be subsetted by arbitrary indices.
(If Frame can be subsetted by a:b, it is likely to be a sequence and thus a subset may be made by row index?)
df <- data.frame(group = sample(c("a", "b"), 20, replace = T),
val = rnorm(20))
# closure - returns a function that accepts from and to
subsetter <- function(from, to) {
function(x) {
x[from:to, ]
}
}
# from and to are specified
sub1 <- subsetter(2, 4)
sub2 <- subsetter(1, 5)
# data is split from to to
sub1(df)
#group val
#2 a 0.5518802
#3 b 1.5955093
#4 a -0.8132578
sub2(df)
# group val
#1 b 0.4780080
#2 a 0.5518802
#3 b 1.5955093
#4 a -0.8132578
#5 b 0.4449554
I am struggling a bit with a probably fairly simple task. I wanted to create a function that has arguments of dataframe(df), column names of dataframe(T and R), value of the selected column of dataframe(a and b). I know that the function reads the dataframe. but , I don't know how the columns are selected. I'm getting an error.
fun <- function(df,T,a,R,b)
{
col <- ds[c("x","y")]
omit <- na.omit(col)
data1 <- omit[omit$x == 'a',]
data2 <- omit[omit$x == 'b',]
nrow(data2)/nrow(data1)
}
fun(jugs,Place,UK,Price,10)
I'm new to r language. So, please help me.
There are several errors you're making.
col <- ds[c("x","y")]
What are x and y? Presumably they're arguments that you're passing, but you specify T and R in your function, not x and y.
data1 <- omit[omit$x == 'a',]
data2 <- omit[omit$x == 'b',]
Again, presumably, you want a and b to be arguments you passed to the function, but you specified 'a' and 'b' which are specific, not general arguments. Also, I assume that second "omit$x" should be "omit$y" (or vice versa). And actually, since you just made this into a new data frame with two columns, you can just use the column index.
nrow(data2)/nrow(data1)
You should print this line, or return it. Either one should suffice.
fun(jugs,Place,UK,Price,10)
Finally, you should use quotes on Place, UK, and Price, at least the way I've done it.
fun <- function(df, col1, val1, col2, val2){
new_cols <- df[,c(col1, col2)]
omit <- na.omit(new_cols)
data1 <- omit[omit[,1] == val1,]
data2 <- omit[omit[,2] == val2,]
print(nrow(data2)/nrow(data1))
}
fun(jugs, "Place", "UK", "Price", 10)
And if I understand what you're trying to do, it may be easier to avoid creating multiple dataframes that you don't need and just use counts instead.
fun <- function(df, col1, val1, col2, val2){
new_cols <- df[,c(col1, col2)]
omit <- na.omit(new_cols)
n1 <- sum(omit[,1] == val1)
n2 <- sum(omit[,2] == val2)
print(n2/n1)
}
fun(jugs, "Place", "UK", "Price", 10)
I would write this function as follows:
fun <- function(df,T,a,R,b) {
data <- na.omit(df[c(T,R)]);
sum(data[[R]]==b)/sum(data[[T]]==a);
};
As you can see, you can combine the first two lines into one, because in your code col was not reused anywhere. Secondly, since you only care about the number of rows of the two subsets of the intermediate data.frame, you don't actually need to construct those two data.frames; instead, you can just compute the logical vectors that result from the two comparisons, and then call sum() on those logical vectors, which naturally treats FALSE as 0 and TRUE as 1.
Demo:
fun <- function(df,T,a,R,b) { data <- na.omit(df[c(T,R)]); sum(data[[R]]==b)/sum(data[[T]]==a); };
df <- data.frame(place=c(rep(c('p1','p2'),each=4),NA,NA), price=c(10,10,20,NA,20,20,20,NA,20,20), stringsAsFactors=F );
df;
## place price
## 1 p1 10
## 2 p1 10
## 3 p1 20
## 4 p1 NA
## 5 p2 20
## 6 p2 20
## 7 p2 20
## 8 p2 NA
## 9 <NA> 20
## 10 <NA> 20
fun(df,'place','p1','price',20);
## [1] 1.333333