How to read factors' levels right in R? - r

Question

I have a big csv file that has 51993 rows and 18 columns. Here is part of the table:
head(ddd)
country.of.birth age sex X2000 X2001 X2002 X2003 X2004 X2005 X2006 X2007
Afghanistan 0 men 0 0 1 2 2 0 1 1
Afghanistan 0 women 1 1 0 0 1 0 0 0
Afghanistan 1 men 0 2 5 2 3 4 1 1
Afghanistan 1 women 4 1 4 2 3 2 3 2
Afghanistan 2 men 5 0 8 7 7 3 5 3
Afghanistan 2 women 4 8 3 9 4 4 4 3
In the main csv file, the columns are: Country of Birth, age, sex, and then years from 2000 to 2014. My questions are why does R put X before each year number?
When I used the str() function, I got:
> str(ddd)
'data.frame': 15264 obs. of 18 variables:
$ country.of.birth: Factor w/ 261 levels "0","1","10","103",..: 51 51 51 51 51 51 51 51 51 51 ...
$ age : Factor w/ 38 levels "","0 ","1 ","10 ",..: 2 2 3 3 14 14 17 17 20 20 ...
$ sex : Factor w/ 39 levels "","0 ","1 ","10 ",..: 38 39 38 39 38 39 38 39 38 39 ...
$ X2000 : Factor w/ 786 levels "","0","1","10",..: 2 3 2 478 555 478 92 4 205 716 ...
$ X2001 : int 0 1 2 1 0 8 11 8 26 19 ...
$ X2002 : int 1 0 5 4 8 3 13 18 22 15 ...
$ X2003 : int 2 0 2 2 7 9 15 13 23 33 ...
$ X2004 : int 2 1 3 3 7 4 11 15 21 22 ...
$ X2005 : int 0 0 4 2 3 4 10 6 13 16 ...
$ X2006 : int 1 0 1 3 5 4 8 13 20 10 ...
$ X2007 : int 1 0 1 2 3 3 6 7 9 17 ...
$ X2008 : int 0 0 2 0 4 5 4 6 8 9 ...
$ X2009 : int 0 1 1 4 7 3 9 10 11 12 ...
$ X2010 : int 1 1 6 4 8 10 17 10 21 16 ...
$ X2011 : int 0 5 9 6 21 18 16 27 34 24 ...
$ X2012 : int 3 5 5 16 30 22 44 48 46 49 ...
$ X2013 : int 3 0 12 19 24 34 54 46 76 71 ...
$ X2014 : int 2 3 15 3 21 29 37 48 64 62 ...
As you notice, sex is factor but with 39 levels whereas it has only two values (men) and (women). Also, year 2000 (X2000 in the table) is a factor with 786 levels and it should have been read as an "int". Why did R read the observable "sex" with this large number of levels and why did it read year 2000 as a factor while it read the others as int (as is the case)?
Edit:
The age column has values of the form: 20-24, 25-30, ... till 85-90. and another category as 90+

Answer 1

X is put in front of the column names because R doesn't allow first character of the column name to be a number (try data.frame(a = 1:10, "3" = runif(10)).
Age is a factor because you have bins and what you observe is expected behavior. R doesn't handle intervals as numeric, but as factor.
Sex variable is weird and given currently available data, I would say the variable represents something other than sex, in at least part of the dataset. Has the dataset been stitched together? Perhaps there was a mistake in copy/pasting. See levels(ddd$sex) to disentangle all possible levels.

Answer 2

The default behaviour of read.table and its related functions is to make all column names syntactically valid. This means that they can be used without quoting after the $ operator. However, this behaviour can be changed using the check.names = FALSE parameter. This will mean you end up with columns called 2000 etc. To then use those columns with $ they will need to be backquoted, e.g.
ddd$`2000`
The same will be true if you want to use these columns with non-standard evaluation, e.g.
ggplot(ddd, aes(x = sex, y = `2000`)) + geom_boxplot()
For the sex column, there must be entries in the column further down that have numbers in. Check your original
data.
For age, you have trailing spaces in your age column. Either remove these outside R, or you could do something like this:
ddd$age <- as.numeric(sub(" +$", "", as.character(ddd$age)))
For the 2000 column, it's not clear from your str output why it's been read as a factor. By default, empty strings should be regarded as NA and so shouldn't affect the class. You could try (assuming you're now using check.names = FALSE):
as.character(ddd$`2000`)[is.na(as.numeric(as.character(ddd$`2000`))) & ddd$`2000` != ""]
This should print out any elements of the column which are non-blank and non-numeric. It may again be a trailing space issue.