[R] help with as.numeric

[Marc Schwartz]

On May 15, 2009, at 6:57 AM, deanj2k wrote:

>
> hi everyone, wondering if you could help me with a novice problem.   
> I have a
> data frame called subjects with a height and weight variable and  
> want to
> calculate a bmi variable from the two.  i have tried:
>
> attach(subjects)
> bmi <- (weight)/((height/100)^2)
>
> but it comes up with the error:
> Warning messages:
> 1: In Ops.factor(height, 100) : / not meaningful for factors
> 2: In Ops.factor((weight), ((height/100)^2)) :
>  / not meaningful for factors
>
> I presume that this means the vectors height and weight are not in  
> numeric
> form (confirmed by is.numeric) so i changed the code to:
>
> bmi <- (as.numeric(weight))/((as.numeric(height)/100)^2)
>
> but this just comes up with a result which doesnt make sense i.e.  
> numbers
> such as 40000 within bmi vector.  Ive looked at
> as.numeric(height)/as.numeric(weight) and these numbers just arnt  
> the same
> as height/weight which is the reason for the incorrect bmi.  Cant  
> anyone
> tell me where I am going wrong?  Its quiet frustrating because I cant
> understand why a function claiming to convert to numeric would come  
> up with
> such a bizarre result.


That 'height' is a factor suggests that you imported the data using  
one of the read.table() family of functions and that there are non- 
numeric characters in at least one of the entries in that column.

Since 'height' is a factor, if you use as.numeric(), you will get  
numeric values returned that are the factor level numeric codes and  
not the expected numeric values. That is why you are getting bad  
values for BMI.

See:

   http://cran.r-project.org/doc/FAQ/R-FAQ.html#How-do-I-convert-factors-to-numeric_003f


If you use something like:

   grep("[^0-9\\.]", height, value = TRUE)

that should show you where you have non-numeric values in the 'height'  
column. That is, entries for 'height' that contain characters other  
than numeric or a decimal. Foe example:

height <- factor(c(seq(0, 1, 0.1), "1,10", letters[1:5]))

 > height
  [1] 0    0.1  0.2  0.3  0.4  0.5  0.6  0.7  0.8  0.9  1    1,10 a     
b    c    d    e
Levels: 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1,10 a b c d e

 > as.numeric(height)
  [1]  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15 16 17

 > grep("[^0-9\\.]", height, value = TRUE)
[1] "1,10" "a"    "b"    "c"    "d"    "e"


I would also check the 'weight' column for the same reasons, to be  
sure that you don't have bad data there. Another approach would be to  
use:

   str(subjects)

which will give you a sense of the data types for each column in your  
data frame. Review each column and take note of any columns that  
should be numeric, but are factors.

See ?str, ?grep and ?regex for more information. You might also want  
to look at ?type.convert, which is the function used by the  
read.table() family of functions to determine the data types for each  
column during import.

HTH,

Marc Schwartz