Chapter 2

Getting Data into R

In the following chapter we address entering data into R and organising it as scalars (single values), vectors, matrices, data frames, or lists. We also demonstrate importing data from Excel, ascii files, databases, and other statistical programs.

2.1 First Steps in R 2.1.1 Typing in Small Datasets
We begin by working with an amount of data that is small enough to type into R. We use a dataset (unpublished data, Chris Elphick, University of Connecticut) containing seven body measurements taken from approximately 1100 saltmarsh sharp-tailed sparrows (Ammodramus caudacutus) (e.g., size of the head and wings, tarsus length, weight, etc.). For our purposes we use only four morphometric variables of eight birds (Table 2.1).
Table 2.1 Morphometric measurements of eight birds. The symbol NA stands for a missing value. The measured variables are the lengths of the wing (measured as the wing chord), leg (a standard measure of the tarsus), head (from the bill tip to the back of the skull), and weight. Wingcrd Tarsus Head Wt 59 55 53.5 55 52.5 57.5 53 55 22.3 19.7 20.8 20.3 20.8 21.5 20.6 21.5 31.2 30.4 30.6 30.3 30.3 30.8 32.5 NA 9.5 13.8 14.8 15.2 15.5 15.6 15.6 15.7

The simplest, albeit laborious, method of entering the data into R is to type it in as scalars (variables containing a single value). For the first five observations of wing length, we could type:
A.F. Zuur et al., A Beginner’s Guide to R, Use R, DOI 10.1007/978-0-387-93837-0_2, Ó Springer ScienceþBusiness Media, LLC 2009 29

30

2 Getting Data into R

> > > > >

a b c d e

Wing1 Wing2 Wing3 Wing4 Wing5 sqrt(Wing1) 2 * Wing1 Wing1 + Wing2 Wing1 + Wing2 + Wing3 + Wing4 + Wing5 (Wing1 + Wing2 + Wing3 + Wing4 + Wing5) / 5

Although R performs the calculations, it does not store the results. It is perhaps better to define new variables: > > > > > SQ.wing1 sum(Wingcrd) [1] 440.5 Obviously, we can also store the sum in a new variable > S.win S.win [1] 440.5 Again, the dot is part of the variable name. Now, enter the data for the other three variables from Table 2.1 into R. It is laborious, but typing the following code into an editor, then copying and pasting it into R does the job. > Tarsus Head Wt sum(Head) [1] NA You will get the same result with the mean, min, max, and many other functions. To understand why we get NA for the sum of the head values, type ?sum. The following is relevant text from the sum help file. ... sum(..., na.rm = FALSE) ... If na.rm is FALSE, an NA value in any of the arguments will cause a value of NA to be returned, otherwise NA values are ignored. ... Apparently, the default ‘‘na.rm = FALSE’’ option causes the R function sum to return an NA if there is a missing value in the vector (rm refers to remove). To avoid this, use ‘‘na.rm = TRUE’’ > sum(Head, na.rm = TRUE) [1] 216.1 Now, the sum of the seven values is returned. The same can be done for the mean, min, max, and median functions. On most computers, you can also use na.rm = T instead of na.rm = TRUE. However, because we have been confronted with classroom PCs running identical R versions on the same operating system, and a few computers give an error message with the na.rm = T option, we advise using na.rm = TRUE. You should always read the help file for any function before use to ensure that you know how it deals with missing values. Some functions use na.rm, some use na.action, and yet others use a different syntax. It is nearly impossible to memorise how all functions treat missing values. Summarising, we have entered data for four variables, and have applied simple functions such as mean, min, max, and so on.We now discuss methods of combining the data of these four variables: (1) the c, cbind, and rbind functions; (2) the matrix and vector functions; (3) data frames; and (4) lists. Do Exercise 1 in Section 2.4 in the use of the c and sum functions.

34

2 Getting Data into R

2.1.3 Combining Variables with the c, cbind, and rbind Functions
We have four columns of data, each containing observations of eight birds. The variables are labelled Wingcrd, Tarsus, Head, and Wt. The c function was used to concatenate the eight values. In the same way as the eight values were concatenated, so can we concatenate the variables containing the values using: > BirdData BirdData [1] 59.0 55.0 [10] 19.7 20.8 [19] 30.6 30.3 [28] 15.2 15.5

53.5 20.3 30.3 15.6

55.0 20.8 30.8 15.6

52.5 57.5 53.0 55.0 22.3 21.5 20.6 21.5 31.2 30.4 32.5 NA 9.5 13.8 14.8 15.7

BirdData is a single vector of length 32 (4 Â 8). The numbers [1], [10], [19], and [28] are the index numbers of the first element on a new line. On your computer they may be different due to a different screen size. There is no need to pay any attention to these numbers yet. R produces all 32 observations, including the missing value, as a single vector, because it does not distinguish values of the different variables (the first 8 observations are of the variable Wingcrd, the second 8 from Tarsus, etc.) . To counteract this we can make a vector of length 32, call it Id (for ‘‘identity’’), and give it the following values. > Id Id [1] 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 3 3 3 3 3 3 3 [24] 3 4 4 4 4 4 4 4 4 Because R can now put more digits on a line, as compared to in BirdData, only the indices [1] and [24] are produced. These indices are completely irrelevant for the moment. The variable Id can be used to indicate that all observations with a similar Id value belong to the same morphometric variable. However, creating such a vector is time consuming for larger datasets, and, fortunately, R has functions to simplify this process. What we need is a function that repeats the values 1 –4, each eight times:

2.1 First Steps in R

35

> Id Id [1] 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 3 3 3 3 3 3 3 [24] 3 4 4 4 4 4 4 4 4 This produces the same long vector of numbers as above. The rep designation stands for repeat. The command can be further simplified by using: > Id Id [1] 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 3 3 3 3 3 3 3 [24] 3 4 4 4 4 4 4 4 4 Again, we get the same result. To see what the 1 : 4 command does, type into R: > 1 : 4 It gives [1] 1 2 3 4 So the : operator does not indicate division (as is the case with some other packages). You can also use the seq function for this purpose. For example, the command > a a creates the same sequence from 1 to 4, [1] 1 2 3 4 So for the bird data, we could also use: > a rep(a, each = 8) [1] 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 3 3 3 3 3 3 3 [24] 3 4 4 4 4 4 4 4 4 Each of the digits in ‘‘a’’ is repeated eight times by the rep function. At this stage you may well be of the opinion that in considering so many different options we are making things needlessly complicated. However, some functions in R need the data as presented in Table 2.1 (e.g, the multivariate analysis function for principal component analysis or multidimensional scaling), whereas the organisation of data into a single long vector, with an extra variable to identify the groups of observations (Id in this case), is needed for other functions such as the t-test, one-way anova, linear regression, and also for some graphing tools such as the xyplot in the lattice package (see Chapter 8). Therefore, fluency with the rep function can save a lot of time.

36

2 Getting Data into R

So far, we have only concatenated numbers. But suppose we want to create a vector ‘‘Id’’ of length 32 that contains the word ‘‘Wingcrd’’ 8 times, the word ‘‘Tarsus’’ 8 times, and so on.We can create a new variable called VarNames, containing the four morphometric variable designations. Once we have created it, we use the rep function to create the requested vector: > VarNames VarNames [1] "Wingcrd" "Tarsus" "Head" "Wt" Note that these are names, not the variables with the data values. Finally, we need: > Id2 Id2 [1] "Wingcrd" "Wingcrd" "Wingcrd" [5] "Wingcrd" "Wingcrd" "Wingcrd" [9] "Tarsus" "Tarsus" "Tarsus" [13] "Tarsus" "Tarsus" "Tarsus" [17] "Head" "Head" "Head" [21] "Head" "Head" "Head" [25] "Wt" "Wt" "Wt" [29] "Wt" "Wt" "Wt"

"Wingcrd" "Wingcrd" "Tarsus" "Tarsus" "Head" "Head" "Wt" "Wt"

Id2 is a string of characters with the names in the requested order. The difference between Id and Id2 is just a matter of labelling. Note that you should not forget the "each =" notation. To see what happens if it is omitted, try typing: > rep(VarNames, 8) [1] "Wingcrd" "Tarsus" [5] "Wingcrd" "Tarsus" [9] "Wingcrd" "Tarsus" [13] "Wingcrd" "Tarsus" [17] "Wingcrd" "Tarsus" [21] "Wingcrd" "Tarsus" [25] "Wingcrd" "Tarsus" [29] "Wingcrd" "Tarsus"

"Head" "Head" "Head" "Head" "Head" "Head" "Head" "Head"

"Wt" "Wt" "Wt" "Wt" "Wt" "Wt" "Wt" "Wt"

It will produce a repetition of the entire vector VarNames with the four variable names listed eight times, not what we want in this case. The c function is a way of combining data or variables. Another option is the cbind function. It combines the variables in such a way that the output contains the original variables in columns. For example, the output of the cbind function below is stored in Z. If we type Z and press enter, it shows the values in columns:

2.1 First Steps in R

37

> Z Z Wingcrd Tarsus [1,] 59.0 22.3 [2,] 55.0 19.7 [3,] 53.5 20.8 [4,] 55.0 20.3 [5,] 52.5 20.8 [6,] 57.5 21.5 [7,] 53.0 20.6 [8,] 55.0 21.5

Tarsus, Head, Wt) Head 31.2 30.4 30.6 30.3 30.3 30.8 32.5 NA Wt 9.5 13.8 14.8 15.2 15.5 15.6 15.6 15.7

The data must be in this format if we are to apply, for example, principal component analysis. Suppose you want to access some elements of Z, for instance, the data in the first column. This is done with the command Z [, 1]: > Z[, 1] [1] 59.0 55.0 53.5 55.0 52.5 57.5 53.0 55.0 Alternatively, use > Z[1 : 8, 1] [1] 59.0 55.0 53.5 55.0 52.5 57.5 53.0 55.0 It gives the same result. The second row is given by Z [2,] : > Z[2, ] Wingcrd 55.0 Tarsus 19.7 Head 30.4 Wt 13.8

Alternatively, you can use: > Z[2, 1:4] Wingcrd Tarsus 55.0 19.7 Head 30.4 Wt 13.8

The following commands are all valid. > > > > > > > Z[1, 1] Z[, 2 : 3] X > > > > > > > > W > > > > > >

x1