Importing a data frame vs creating one in R - r

Question

I am trying to create a specialized summary 'matrix' for my supervisor, and would like R to export it in a clean, readable form. As such, I am creating it from scratch basically, to tailor it to our project. My problem is I can't figure out how to get a created data frame to behave like an imported one, specifically headers.
I am most comfortable dealing with imported data frames with headers, and calling specific rows by name instead of column number:
iris$Sepal.Length
with(iris,Sepal.Length)
iris['Sepal.Length']
Now, if I want to create a data frame (or matrix, I'm not entirely sure what the difference is), I have tried the following:
groups<-c("Group 1", "Group 2")
factors<-c("Fac 1", "Fac 2", "Fac 3","Fac 4", "Fac 5")
x<-1:10
y<-11:20
z<-21-30
data<-cbind(groups, factors, x, y, z)
names(data) #returns NULL
data$x #clearly doesn't return the column 'x' since the matrix 'data' has no names
data<-data.frame(cbind(groups, factors, x, y, z))
names(data) #confirms that there are header names
So, I have created a data frame that has the columns x, y and z, but in reality I don't have a premade column to start off with. If I knew how many rows of data there would be I could simply do:
data<-data.frame(1:10)
data$x<-x
data$y<-y
data$z<-z
I tried creating an empty data frame, but it is one element big, and if I try to append a vector to it (of any length greater than 1), I get an error:
data<-data.frame(0)
data$x<-x #returns an error
My best guess at what to do is to pass through the data once to find out how long many rows of data I will have (there are several factor levels, and the summary matrix will have a row for each possible combination of factors). Then I can get the data frame started with a simple:
data<-data.frame(length(n)) #where n would be how many rows of data I would have
And follow through by creating individual vectors for each summary statistic I want and appending it to the data frame with ~$~.
Another solution I tried to play with was creating a matrix and filling in each element as I calculate it within a loop. I know the apply family is better than a loop, but to make my summary table tailored to my needs I would need to run an apply function then try to pull the individual data:
means<-with(iris,tapply(iris[,4],Species,mean))
means[1] #This returns the species and the mean petal width. What I need is the numeric part of this, as I will have my own headers, or possibly a separate summary table for each species.
I'm not sure if extracting the numerical information from the apply output is better / any easier than simply constructing my own loop to calculate the required statistics. It would be a nested loop that would first sort by group (2 runs), then an internal loop that would run by factors (5 runs) for a total of 10 runs through the data. I was thinking of creating an empty martix, and simply saving the data in the appropriate cell when it is calculated. My problem, again, is calling a specific row in a matrix. I have tried:
m<-matrix(0,ncol=5)
m[1,1]<-'Groups'
m[1,2]<-'Factors'
m[1,3]<-'Mean.x'
m[1,4]<-'Mean.y'
m[1,5]<-'Mean.z'
names(m) #Returns NULL
My desired output would look like:
Groups Factors Mean.x Mean.y Mean.z
Group 1 Fac 1
Group 1 Fac 2
Group 1 Fac 3
Etc, for all combinations of groups and factors.

Answer 1

You can use ddply from plyr package for that: assume your original data frame is mydata and your new data frame where you store the result is newdata:
library(plyr)
newdata<-ddply(mydata,.(Groups,Factors),summarize,mean.x=mean(x),mean.y=mean(y),mean.z=mean(z))
Example: mydata<-iris
> newdata<-ddply(mydata,.(Species),colwise(mean))
> newdata
Species Sepal.Length Sepal.Width Petal.Length Petal.Width
1 setosa 5.006 3.428 1.462 0.246
2 versicolor 5.936 2.770 4.260 1.326
3 virginica 6.588 2.974 5.552 2.026

Answer 2

I think this is what you're looking for, but I'm a little confused by your question in general. This basically will give you a pivot table of the means in each column x,y, and z grouped by the columns 'groups' and 'factors'
aggregate(.~groups+factors, data=data, FUN="mean")
groups factors x y z
1 Group 1 Fac 1 1 1 1
2 Group 2 Fac 1 7 6 1
3 Group 1 Fac 2 8 7 1
4 Group 2 Fac 2 3 2 1
5 Group 1 Fac 3 4 3 1
6 Group 2 Fac 3 9 8 1
7 Group 1 Fac 4 10 9 1
8 Group 2 Fac 4 5 4 1
9 Group 1 Fac 5 6 5 1
10 Group 2 Fac 5 2 10 1
or with the iris data grouped by Species:
aggregate(.~Species, data=iris, FUN="mean")
Species Sepal.Length Sepal.Width Petal.Length Petal.Width
1 setosa 5.006 3.428 1.462 0.246
2 versicolor 5.936 2.770 4.260 1.326
3 virginica 6.588 2.974 5.552 2.026
UPDATE: To only calculate the mean of certain columns, you can either pass only the apropriate columns of your dataset to the aggregate function (perhaps calling subset) or modify the formula like this:
aggregate(cbind(Sepal.Length,Sepal.Width)~Species, data=iris, FUN="mean")

Answer 3

I am not entirely sure if that's what you are looking for but there are several options to add “stuff” to data frames:
To add a variable, just type data$newname <- NA (no need to know the length of the data frame or pass a vector, all rows will be filled with NA)
To append data use rbind (the data you are adding should be another data frame with the same variables)
To fix your example, first create an empty data frame and append data as it comes:
data <- data.frame(x=numeric())
data <- rbind(data, data.frame(x))
The previous example had only one variable (x) but you can also define a data frame with several variables and no rows:
data <- data.frame(x=numeric(),
y=numeric(),
a=character(),
b=factor(levels=c("Factor 1", "Factor 2")))
You don't need to know how many rows you will have, but the data you are adding needs to have the same structure. If that's not the case, you need to create columns with missing values in both data frames as needed, e.g.
data1 <- data.frame(x=1:10, y=1)
data2 <- data.frame(y=2, z=100:110)
rbind(data1, data2) # Error
data1$z <- NA
data2$x <- NA
rbind(data1, data2) # Now it works