Re: st: Data management question

[Michael Hanson <mshanson@mac.com>]

I agree with Nick -- who beat me to the punch on this very  
recommendation.  Some slight additional value-added:  -trim()- will  
not remove spaces within a string, only at the beginning and end, and  
Nick's suggestion below will count those spaces as if they too were  
"X"s.  To make your count variable (which I call "n") robust to such a  
possibility, you could instead use:

gen n = length(subinstr(x, " ", "", .))

where "x" is your original string variable.  Ideally, you can  
"robustify" this approach further by using regular expression  
functions, but I have not yet determined a clean way to remove _all_  
possible non-letter characters ([:space:] characters in regex) -- at  
best, -regexr()- only replaces the _first_ occurrence of a [:space:]  
character with a blank, not all of them.  Suggestions on this point  
would be welcome.

Also, if you want a truly robust solution, you should decide whether  
non-"X" letter characters should be counted or not.  Again, clever use  
of a regex expression ought to help with this, but personally I have  
not figured out how to get Stata to deliver on this potential.

Hope this helps,
Mike


On Jul 16, 2009, at 8:39 AM, Nick Cox wrote:

> As a footnote, observe that if the issue is counting "X" in values of
> -kisses- such as "XX", "XXXXX", etc. then
>
> gen nkisses = length(kisses)
>
> will be a more direct and efficient solution so long as no other
> characters are observed. -length(trim(kisses))- will protect against
> accidental leading and trailing spaces.
>
> Nick
> n.j.cox@durham.ac.uk
>
> Susan Olivia
>
> Thanks Tirthankar,
>
> Using Nick Cox's command is way more efficient. My earlier
> attempts were very inefficient.
>
> Tirthankar Chakravarty
>
> > There are probably many ways of doing this, but here is
> > way using Nick Cox's -egenmore- (SSC, Nick Winter is
> > attributed as the author of the -noccur()- function used
> > here) package:
>
> > clear*
> > set obs 100
> > g crosses = " "
> > local cross "x"
> > forv i=1/ 100 {
> >       qui: replace crosses="`cross'" in `i'
> >       local cross "`cross'x"
> > }
> > // ssc install egemore, replace
> > egen noccur = noccur(crosses), string("x")
> > su noccur
>
> > On Thu, Jul 16, 2009 at 2:26 AM, Susan
> Olivia<olivia@primal.ucdavis.edu>
>
> > > I have a variable (say number of days) and is a string
> > > variable. This variable is represented by XXXX
> > > (basically the number of X denotes the number of days).
> > >
> > > I would like to create a numeric value for this variable
> > > (i.e. 4 crosses = 4) . Is there a way I can easily do
> > > this in Stata? I tried the 'encode' and 'destring'
> > > commands, but these commands didn't do what I after.

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/