how to chop off the last 2 or 3 digits

[Jahan zee]

I have variable name CUSIP which has either 6, 8 or 9 character length (Alpha numeric or Numeric), I want to chop off the last 2 or 3 (characters or digits) which has length of 8 or 9 characters, in order to make all of them to 6 digit.

[Jesse Wursten]

Try

gen cusip_short = substr(CUSIP, 1, 6)

[Jahan zee]

Try

gen cusip_short = substr(CUSIP, 1, 6)

It do not work.

[Jesse Wursten]

What error message do you get? Is your variable actually a string?

[Jahan zee]

No error msg but when I check the values those are still in 8 or 9 Character long. Here is my data. first 58 values are 6 character long so generated dataex for 75 observations here

* Example generated by -dataex-. To install: ssc install dataex
clear
input str17 cusip int fyear
"360206"    1990
"360206"    1991
"360206"    1992
"360206"    1993
"360206"    1994
"360206"    1995
"360206"    1996
"360206"    1997
"360206"    1998
"360206"    1999
"360206"    2000
"360206"    2001
"360206"    2002
"360206"    2003
"360206"    2004
"360206"    2005
"360206"    2006
"361105"    1988
"361105"    1989
"361105"    1990
"361105"    1991
"361105"    1992
"361105"    1993
"361105"    1994
"361105"    1995
"361105"    1996
"361105"    1997
"361105"    1998
"361105"    1999
"361105"    2000
"361105"    2001
"361105"    2002
"361105"    2003
"361105"    2004
"361105"    2005
"375204"    1997
"375204"    1998
"375204"    1999
"375204"    2000
"375204"    2001
"375204"    2002
"375204"    2003
"375204"    2004
"375204"    2005
"375204"    2006
"781104"    1989
"781104"    1990
"781104"    1991
"781104"    1992
"781104"    1993
"781104"    1994
"782102"    1992
"782102"    1993
"782102"    1994
"782102"    1995
"782102"    1996
"782102"    1997
"00086T103" 1995
"00086T103" 1996
"00086T103" 1997
"00086T103" 1998
"00086T103" 1999
"00086T103" 2000
"00086T103" 2001
"00086T103" 2002
"00086T103" 2003
"00086T103" 2004
"00086T103" 2005
"00086T103" 2006
"872309"    1989
"872309"    1990
"872309"    1991
"872309"    1992
"872309"    1993
"872309"    1994
end

[Nick Cox]

clear
input long whatever
123456
1234567
12345678
end

gen whatev = ///
cond(whatever < 1e6, whatever, cond(whatever  < 1e7, floor(whatever/10), floor(whatever/100)))

list

     +-------------------+
     | whatever   whatev |
     |-------------------|
  1. |   123456   123456 |
  2. |  1234567   123456 |
  3. | 12345678   123456 |
     +-------------------+

.

or

gen WHATEV = real(substr(string(whatever, "%9.0f"), 1, 6))

[Jesse Wursten]

Did you look at the cusip_short variable? Because if it did not give an error message, I do not see how the new variable could be longer than 6 characters. You can use replace CUSIP = ... instead of gen cusip_short = ... if you want to replace the original variable.

Nick Cox
The original variable is already a string, as there are "letters" in there, eg "00086T103".

[Nick Cox]

Jesse: Quite. I didn't see #5 when I was posting #6. I can't see why your solution in #2 is not bang on.

[William Lisowski]

There are two possibilities here.

1) In posts #1 and #2 the variable name was CUSIP, but in the sample data in post #5 the variable name is cusip. However, applying the code from #2 to the data from #5 generates an error message, but perhaps this was overlooked.

2) More likely, after applying the code from #2 to the data, the original poster looked at the existing variable cusip rather than the new variable cusip_short generated by the code.

. gen cusip_short = substr(cusip, 1, 6)

. list in 58/69, clean

           cusip   fyear   cusip_~t  
 58.   00086T103    1995     00086T  
 59.   00086T103    1996     00086T  
 60.   00086T103    1997     00086T  
 61.   00086T103    1998     00086T  
 62.   00086T103    1999     00086T  
 63.   00086T103    2000     00086T  
 64.   00086T103    2001     00086T  
 65.   00086T103    2002     00086T  
 66.   00086T103    2003     00086T  
 67.   00086T103    2004     00086T  
 68.   00086T103    2005     00086T  
 69.   00086T103    2006     00086T

[Jahan zee]

There are two possibilities here.

1) In posts #1 and #2 the variable name was CUSIP, but in the sample data in post #5 the variable name is cusip. However, applying the code from #2 to the data from #5 generates an error message, but perhaps this was overlooked.

2) More likely, after applying the code from #2 to the data, the original poster looked at the existing variable cusip rather than the new variable cusip_short generated by the code.

Thanks I understand the problem.