R: integer versus number

Question

This question may have been answered earlier, but I did not see the answer.

I have a data set consisting of numbers and missing values. One row is a percentage. The following is a small set of fake data in which the column names are AA, BB, and CC. The third row in this dataset is the percentage.

AA BB CC 234 432 78 1980 3452 2323 91.1 90 93.3 34 123 45 

In this case, when I read the dataset, AA and CC are numeric, and BB is an integer. I guess somewhere around 90.0 was rounded to 90. If I don't indicate that BB is numeric, can this cause problems with basic arithmetic?

I believe that if dd = 1 and ee = 2, and both are integers, then the C-language says dd / ee = 0, and R says dd / ee = 0.5.

The following is a series of simple mathematical operations that seem to suggest that the answers in R do not change regardless of whether the data is numeric or integer. However, I continue to think that it would be prudent to indicate that all variables are numeric when reading data. Using Google, I found an example or two where the data type really mattered, but not lower.

 aa <- c(1,2,3,4,5,6,7) bb <- 2 str(aa) str(bb) cc <- as.integer(aa) dd <- as.integer(bb) str(cc) str(dd) aa/bb cc/dd aa/dd cc/bb ee <- aa * aa str(ee) sum(ee/2) ff <- cc * cc str(ff) sum(ff/2) gg <- 4.14 hh <- ((aa * aa) * gg) / 2 hh ii <- ((cc * cc) * gg) / 2 ii jj <- (aa * aa) / gg jj kk <- (cc * cc) / gg kk jj == kk mm <- as.integer(1) nn <- as.integer(2) mm/nn 

I think I hope for confidence that this is hardly a problem with simple math, but I suspect that this is possible. I keep thinking that there is a basic programming rule here, but I'm not sure what it is. (I know the concept of double precision.)

Thanks for any tips that are definitely the main issue.

Answer 1

A unit using the / operator will always return "numeric", that is, equivalent to C "double". Numerators and denominators are first forced to numeric, and then division is performed. If you want to use integer division, you can use %/% . If you want to create an integer, you can use trunc or floor , or you can use round(x , 0) , or you can use as.integer. The first second and fourth of these options are equivalent. The round function will still return "numeric" even if the printed view looks integer. I do not think you need to worry while you are satisfied with the "double" / "digital" results. Heck, we even allow division by 0.

Your aa variable has been classified as "numeric", even though it was entered as a whole chain, but you used:

 aa <- 1:8 # sequences are integer class. 

It sounds like you won’t be too surprised. FAQ 7.31