The title of this section sounds almost scary, but it really isn't. R uses what is known as "vector arithmetic," which means that it operates on a whole column of data at once. You saw this in the previous section when I asked for "sqrt(data4)." To explain what I mean, let me go back to the good old days of programming. If you were writing a program in Fortran, you would write something like (It's been a long time, so that code may be slightly off.)
do 25 i = 1 to 20
y(i) = sqrt(x(i))
25
If you were writing in Pascal, one of my favorites, you would write
for i := 1 to 5
do begin
y(i) <- sqrt(x(i)
end; {for}
And for C++ it would be
for (int i = 1; i <= 20; i++) {
y(i) = sqrt(x(i)
}
All of those sets of commands are telling the computer to take each value of x, one at a time, and take its square root. But R doesn't work that way. In one command it tells itself to take the square root of all of the x values. That may not look all that much different, but it is. It is very much faster, especially when you have a lot of values. In the book I do things like take the mean of 100,000 observations. And I can do it in a tiny fraction of a second. Doing it the old way took me 19.5 seconds.
If you have never done any programming, this may not seem like a big thing. But for those who have, it takes a bit of getting used to. Your commands may look a bit funny to you. Suppose that I really want to take each value of x, multiply it by the corresponding value of y, and store the result. For example, if x = 1, 3, 5 and y = 2, 4, 6, then xy will be equal to 2, 12, 30. But I don't have to go through the list pair by pair. I can just say
x <- c(1, 3, 5) y <- c(2, 4, 6) xy <- x*y print(xy) [1] 2 12 30
I think that I can give you a good idea of just what is going on, and what an R program does, by doing some of the homework exercises in Chapter 2. You don't have to memorize all of these commands, but read them through carefully and understand what they are doing. We can deal with specifics later.
I'll begin with reading in the data in Exercise 2.10. These are in a data file named Ex2-10.dat on the Web site. Then we will make a stem-and-leaf display of them. Then I will go on to Ex2.12 and Ex2.13 and draw a histogram (at bottom) and stem-and-leaf display. Finally I will go on and do most of Ex2.14 - Ex2-19. The results, shown on the R console, are given below. If you work this yourself you will see some comments after the "attach" command. Ignore them.
> data <- read.table(file.choose(), header = TRUE) > head(data) Sex Grade 1 1 66 2 1 67 3 1 67 4 1 73 5 1 73 6 1 74 > attach(data) Ignore comments like the following The following object(s) are masked _by_ '.GlobalEnv': etc. > stem(Grade) The decimal point is at the | 66 | 00000 68 | 0 70 | 000 72 | 000000 74 | 0000000 76 | 78 | 0000000000 80 | 00000000000000 82 | 0000000000 84 | 0000000000000 86 | 0000000000000 88 | 0000000000 90 | 00000 92 | 00000 94 | 00000 > stem(Grade[Sex == 1] ) The decimal point is 1 digit(s) to the right of the | 6 | 677 7 | 334 7 | 5999 8 | 012244 8 | 5566778 9 | 23444 9 | 5 > stem(Grade[Sex == 2] ) The decimal point is 1 digit(s) to the right of the | 6 | 778 7 | 00022334 7 | 55558899999 8 | 0111111111112222223344 8 | 5555555666777777888889999 9 | 00001333 9 | 5 > data2 <- read.table(file.choose(), header = TRUE) > head(data2) CaseNum ADDSC Gender Repeat IQ EngL EngG GPA SocProb Dropout 1 1 45 1 0 111 2 3 2.60 0 0 2 2 50 1 0 102 2 3 2.75 0 0 3 3 49 1 0 108 2 4 4.00 0 0 4 4 55 1 0 109 2 2 2.25 0 0 5 5 39 1 0 118 2 3 3.00 0 0 6 6 68 1 1 79 2 2 1.67 0 1 > attach(data2) > hist(GPA, breaks = 25) #breaks control the number of bars# Graph shown at end > stem(ADDSC) The decimal point is 1 digit(s) to the right of the | 2 | 69 3 | 034456679 4 | 000233444445566677888899999 5 | 0000000001122333455677889 6 | 0001223455556899 7 | 0024568 8 | 55 > x <- c(10, 8, 9, 5, 10, 11, 7, 8, 7, 7) > y <- c(9, 9, 5, 3, 8, 4, 6, 6, 5, 2) > print(sumx <- sum(x)) # This stores sumx as well as print it. [1] 82 > print(sumy <- sum(y)) [1] 57 > print(sumx2 <- sum(x*x)) # or x2 <- x*x; sum(x2)) [1] 702 > print(sumx.sq <-sum(x)*sum(x)) # or sum(x^2)) [1] 6724 > print(n <- length(x)) [1] 10 > print(meanx <- sum(x)/n) [1] 8.2 > print(sumy2 <- sum(y*y)) # or y2 <- y*y then sum(y2)) [1] 377 > print(sumy.sq <- sum(y) * sum(y)) # or sum(y^2)) [1] 3249 > print(b <- (sumy2 - sumy.sq/n)/(n-1)) [1] 5.788889 > print(sumxy <- sumx*sumy) [1] 4674 > print(sumxsumy <- sumx*sumy) [1] 4674 > print(covar <- (sumxy - (sumx)*(sumy)/n)/(n-1)) # This is the covariance [1] 467.4 > print(stdev <- sqrt((sumx2 - sumx^2/n)/(n-1))) [1] 1.813529 > print(sd(x)) [1] 1.813529 >
Last revised 8/20/2010
dch