Skip to content

help with as.numeric

3 messages · deanj2k, Nutter, Benjamin, Marc Schwartz

deanj2k
# 
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.
Nutter, Benjamin
# 
as.numeric() doesn't convert factors to the explicit value, nor should
it.  Under what you're expecting, ff you have a factor where the levels
are "Female" and "Male", using as.numeric() wouldn't produce anything
meaningful.

However, as.numeric() does something much smarter.  It converts "Female"
to 1, and "Male" to 2.  More generally, if you have n levels, it will
produce a vector of values between 1 and n.  This is referred to as the
'internal coding.'

If you want to convert your height and bmi variables to their numeric
values, you need to do
This will get you around the internal coding.

-----Original Message-----
From: r-help-bounces at r-project.org [mailto:r-help-bounces at r-project.org]
On Behalf Of deanj2k
Sent: Friday, May 15, 2009 7:58 AM
To: r-help at r-project.org
Subject: [R] help with as.numeric


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.
Marc Schwartz
#
On May 15, 2009, at 6:57 AM, deanj2k wrote:

            

      
      
      
    

        
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