Bioequivalence and Bioavailability Forum

x <- c("Foo", "Bar")
is.character(x)
[1] TRUE
b <- data.frame(x)
is.character(b)
[1] FALSE
is.data.frame(b)
[1] TRUE
print(b)
    x
1 Foo
2 Bar
  So far, so good. What else?
str(b)
'data.frame':   2 obs. of  1 variable:
 $ x: Factor w/ 2 levels "Bar","Foo": 2 1
  Implicitly you are using the default of data.frame() like here:
c <- data.frame(x, stringsAsFactors=TRUE)
  Check #1
identical(c, b)
[1] TRUE
  Check #2
c == b
        x
[1,] TRUE
[2,] TRUE
  See the point?
typeof(b[, 1]) ##aha, integer?
[1] "integer"
  Correct. Factors are always coded as integers internally.
  Check #3
is.factor(b[, 1])
[1] TRUE
  Note also the output of str(b) above where factors are given in lexical order.
  However, since we defined x <- c("Foo", "Bar"), "Bar" gets the level 2
  and "Foo" the level 1 in b.
b[, 1]+1 ##then let me add 1
[1] NA NA
Warning message:
In Ops.factor(b[, 1], 1) : ‘+’ not meaningful for factors
  Well roared, lion!
  On the other hand:
d <- data.frame(x, stringsAsFactors=FALSE)
print(d)
    x
1 Foo
2 Bar
str(d)
'data.frame':   2 obs. of  1 variable:
 $ x: chr  "Foo" "Bar"
typeof(c[, 1])
[1] "character"
  Is this what you expected?
d[, 1]+1
Error in c[, 1] + 1 : non-numeric argument to binary operator
  Sure. Rubbish in, rubbish out.
  Personally I prefer a straight error over a warning producing NAs.