Transform community data into wide-format for vegan package - r

Question

I am trying to analyze some community data with the vegan package. I have my data in the wrong format, and am looking for ways to change the format.
What I have is something like this:
Habitat Species Abundance
1 A 3
2 B 2
3 C 1
1 D 5
2 A 8
3 F 4
And what I think I need is:
Habitat Species A Species B Species C Species D Species D
1 3 0 0 5 0
2 8 ...... etc
3 0
Or is there any other format that vegan can take? I am trying to calculate similarity in species composition between habitats.

Answer 1

The function matrify() in the labdsv package does exactly this for community analyses.
Takes a data.frame in three column form (sample.id, taxon, abundance) and converts it into full matrix form, and then exports it as data.frame with the appropriate row.names and column names.
In other words, it converts your data from long to wide format so that each row represents a sample (in your case "habitat"; sometimes this would be a "plot"), each column represents a species, and each cell shows the abundance of the given cell's species (column) in the given cell's habitat (row).
Example:
dat <- data.frame(Habitat = c('Hab1','Hab1','Hab2','Hab2','Hab2','Hab3','Hab3'),
Species = c('Sp1','Sp2','Sp1','Sp3','Sp4','Sp2','Sp3'),
Abundance = c(1,2,1,3,2,2,1))
print(dat)
Habitat Species Abundance
1 Hab1 Sp1 1
2 Hab1 Sp2 2
3 Hab2 Sp1 1
4 Hab2 Sp3 3
5 Hab2 Sp4 2
6 Hab3 Sp2 2
7 Hab3 Sp3 1
library(labdsv)
matrify(dat)
Sp1 Sp2 Sp3 Sp4
Hab1 1 2 0 0
Hab2 1 0 3 2
Hab3 0 2 1 0
Bonus:
I rewrote matrify many years ago so that it could handle longitudinal community data
Specifically, my matrify2() function creates rows for each plot-year combination (i.e., resampled rows for the same plot) by duplicating plot (or habitat) row monikers and adding a Year column.
Below is the code:
#Create data.frame with PLOT, YEAR, and ABUNDANCE for each SPEC:
#Creates function that can sort the data.frame output by:
#Columns = individual SPECS, #Rows = plot by Year
#Note: Code modified from matrify() function from labdsv package (v. 1.6-1)
matrify2 <- function(data) {
#Data must have columns: plot, SPEC, abundance measure,Year
if (ncol(data) != 4)
stop("data frame must have four column format")
plt <- factor(data[, 1])
spc <- factor(data[, 2])
abu <- data[, 3]
yrs <- factor(data[, 4])
plt.codes <- sort(levels(factor(plt))) ##object with sorted plot numbers
spc.codes <- levels(factor(spc)) ##object with sorted SPEC names
yrs.codes <- sort(levels(factor(yrs))) ##object with sorted sampling Years
taxa <- matrix(0, nrow = length(plt.codes)*length(yrs.codes), ncol = length(spc.codes)) ##Create empty matrix with proper dimensions (unique(plotxYear) by # of SPEC)
plt.list <- rep(plt.codes,length(yrs.codes)) ##Create a list of all the plot numbers (in order of input data) to add as an ID column at end of function
yrs.list <- rep(yrs.codes,each=length(plt.codes)) ##Create a list of all the Year numbers (in order of input data) to add as an ID column at end of function
col <- match(spc, spc.codes) ##object that determines the alphabetical order ranking of each SPEC in the spc.code list
row.plt <- match(plt, plt.codes) ##object that determines the rank order ranking of each plot of the input data in the plt.code list
row.yrs <- match(yrs,yrs.codes) ##object that determines the rank order ranking of each Year of the input data in the yrs.code list
for (i in 1:length(abu)) {
row <- (row.plt[i])+length(plt.codes)*(row.yrs[i]-1) ##Determine row number by assuming each row represents a specific plot & year in an object of rep(plot,each=Year)
if(!is.na(abu[i])) { ##ONly use value if !is.na .. [ignore all is.NA values]
taxa[row, col[i]] <- sum(taxa[row, col[i]], abu[i]) ##Add abundance measure of row i to the proper SPEC column and plot/Year row. Sum across all identical individuals.
}
}
taxa <- data.frame(taxa) ##Convert to data.frame for easier manipulation
taxa <- cbind(plt.list,yrs.list,taxa) ##Add ID columns for plot and Year to each row already representing the abundance of Each SPEC of that given plot/Year.
names(taxa) <- c('Plot','Year',spc.codes)
taxa
}
Example:
dat.y <- data.frame(Habitat = c('Hab1','Hab1','Hab2','Hab2','Hab2','Hab3','Hab3','Hab1','Hab1','Hab2','Hab2','Hab2','Hab3','Hab3'),
Species = c('Sp1','Sp2','Sp1','Sp3','Sp4','Sp2','Sp3','Sp1','Sp2','Sp1','Sp3','Sp4','Sp2','Sp3'),
Abundance = c(1,2,1,3,2,2,1,1,2,1,3,2,2,1),
Year = c(1,1,1,1,1,1,1,2,2,2,2,2,2,2))
print(dat.y)
Habitat Species Abundance Year
1 Hab1 Sp1 1 1
2 Hab1 Sp2 2 1
3 Hab2 Sp1 1 1
4 Hab2 Sp3 3 1
5 Hab2 Sp4 2 1
6 Hab3 Sp2 2 1
7 Hab3 Sp3 1 1
8 Hab1 Sp1 1 2
9 Hab1 Sp2 2 2
10 Hab2 Sp1 1 2
11 Hab2 Sp3 3 2
12 Hab2 Sp4 2 2
13 Hab3 Sp2 2 2
14 Hab3 Sp3 1 2
matrify2(dat.y)
Plot Year Sp1 Sp2 Sp3 Sp4
1 Hab1 1 1 2 0 0
2 Hab2 1 1 0 3 2
3 Hab3 1 0 2 1 0
4 Hab1 2 1 2 0 0
5 Hab2 2 1 0 3 2
6 Hab3 2 0 2 1 0
Also, FYI, you should get to know labdsv according to the vegan documentation:
Together with the labdsv package, the vegan package provides most standard tools of descriptive community analysis.

Answer 2

You probably want to spread your data. For example:
library(tidyr)
mydata %>%
spread(Species, Abundance)

Answer 3

This is what I would so, using dcast:
Create a data sample: cc=data.frame(habitat=c(1,2,3,1,2,3),species=c('a','e','a','e','g','a'), abundance=sample(1:10000,6)).
Output looks like this (Ignore first column as it is an automatic index created by the ouput operation in R. What is important is the columns):
> cc
> habitat species abundance
> 1 1 a 7814
> 2 2 e 7801
> 3 3 a 9510
> 4 1 e 7443
> 5 2 g 2160
> 6 3 a 4026
>
Now melt: m=melt(cc, id.vars=c("habitat","species")). Output:
habitat species variable value
1 1 a abundance 7814
2 2 e abundance 7801
3 3 a abundance 9510
4 1 e abundance 7443
5 2 g abundance 2160
6 3 a abundance 4026
Now reshape: dcast(m,habitat~species,fun.aggregate=mean), which yields:
habitat a e g
1 1 7814 7443 NaN
2 2 NaN 7801 2160
3 3 6768 NaN NaN
More info about reshape here.
Kf