Using detring to remove special characters

[Xinruo Wang]

Hi all,

I'm trying to remove special characters from a string variable. I have two variables. One contains only "-", another contains "-" and "+".
I successfully removed "-" from the first variable. When I try to remove "-" and "+" together, The Stata shows 'option "+ not allowed'. Here is my code: destring var, gen(new) ignore("-", "+"). How can I fix this? Thank you!

[William Lisowski]

Replace

destring var, gen(new) ignore("-", "+")

with

destring var, gen(new) ignore("-" "+")

removing the comma between the two strings to ignore. See the ouput of help desetring for further details on the syntax of the ignore option on the destring command.

[Xinruo Wang]

Hi William,

I tried the code. Now the error is "var contains characters not specified in ignore(); no generate". I don't see any other special characters. How can I find out what the characters are? Thank you!

[Nick Cox]

tab whatever if missing(real(whatever))

will show you values that can't be converted to numeric. See also chartab (SSC).

[William Lisowski]

Building on Nick's advice, I'm concerned about ignoring "-". Consider the following example.

. destring x, ignore("-" "+") generate(y)
x: characters - + removed; y generated as byte
(2 missing values generated)

. list, noobs

  +--------+
  |  x   y |
  |--------|
  |  1   1 |
  | -2   2 |
  | +3   3 |
  |  -   . |
  |  +   . |
  +--------+

. generate xx = x

. replace xx = "" if trim(xx)=="-"
(1 real change made)

. destring xx, ignore("+") generate(z)
xx: character + removed; z generated as byte
(2 missing values generated)

. list, noobs

  +------------------+
  |  x   y   xx    z |
  |------------------|
  |  1   1    1    1 |
  | -2   2   -2   -2 |
  | +3   3   +3    3 |
  |  -   .         . |
  |  +   .    +    . |
  +------------------+

We see that by ignoring "-" the negative 2 was destringed as a positive 2.

Certainly Nick's advice is the place to start. Systematically review your data. Be careful about ignoring non-numeric characters that are in fact part of numbers.