Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, statalist.org is already up and running.
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: st: Merging files based on name and year
Joseph Monte <email@example.com>
Re: st: Merging files based on name and year
Sun, 4 Sep 2011 09:11:30 +0100
Thanks for the help. It worked well. Unfortunately, I'm stuck at the
final stage where I need to take the average score for each name
conditional on a score being available for each name. Unfortunately, I
do have missing observations.
. input str1 name score
1. "A" 7
2. "A" 8
3. "A" .
4. "B" 6
5. "B" 7
6. "C" 5
. egen meanscore = mean(score), by(name)
| name score meanscore |
1. | A 8 7.5 |
2. | A 7 7.5 |
3. | A . 7.5 |
4. | B 7 6.5 |
5. | B 6 6.5 |
6. | C 5 5 |
I want meanscore not to be calculated for A since there is a missing
observation (i.e. there should be 3 blanks instead of 7.5). How do I
get around this issue? I tried adapting the code from the link below
but was not successful.
On Thu, Aug 25, 2011 at 9:52 AM, Nick Cox <firstname.lastname@example.org> wrote:
> I don't think regex is the only approach here. You could consider
> using -split-.
> A strategy here is to insert new parsing characters yourself. For
> example suppose that ; is not used, which you can check by
> assert strpos(name, ";") == 0
> Then put ; after each terminal element such as "Inc" (there's probably
> jargon I don't know)
> clonevar work = name
> replace work = subinstr(work, "Inc", "Inc;", .)
> replace work = subinstr(work, "LLC", "LLC;", .)
> replace work = subinstr(work, "Corp", "Corp;", .)
> and so on.
> You can get all the terminal elements from your file with just
> individual names. It is the last word (word(,-1)) of the company name.
> You can put that into a new variable and -tab- the results.
> You may need to fix exceptions, which will be shown by the tabulation above.
> Then -split- on ; and then -reshape-.
> Sometimes a very primitive approach like this is much quicker than
> spending hours trying to do it a cleverer way. (If someone were
> exceptionally fluent with regular expressions that wouldn't be true.)
> When regex works it can be a spectacular solution but with many messy
> problems it is often a very long way round.
> On Thu, Aug 25, 2011 at 9:24 AM, Joseph Monte <email@example.com> wrote:
>> The file below is the master file containing names of companies and
>> years. ABC Inc is one company, XYZ Corp is another company, PNG LLC is
>> a third company. I have a total of 1100 different companies. As shown
>> below, sometimes two or more companies are listed in the same field
>> (there are a maximum of 5 companies listed in the same field). The
>> year column has only one year for each observation. There are a total
>> of 800 observations in this file.
>> name year
>> ABC Inc 1986
>> XYZ Corp 1994
>> ABC Inc XYZ Corp 2001
>> PNG LLC 2005
>> XYZ Corp PNG LLC 2007
>> I have a second file with data in the following format. The 1100
>> companies are listed as shown below. YR8084 means the years 1980-1984,
>> YR8591 means the years 1985-1991, and so on. The numbers below each
>> year are scores assigned to each company during a certain period. For
>> example, ABC Inc is assigned a score of 6 during 1980-1984, 7 from
>> 1985-1991, and 9 from 2001-2004. ABC Inc is not assigned a score
>> during other periods. Scores range from 1 to 9 and may be up to 3
>> decimal places.
>> name YR8084 YR8591 YR9200 YR0104 YR0507 YR0809
>> ABC Inc 6 7 9
>> XYZ Corp 2 5 6 6
>> PNG LLC 7 7 7
>> 7 7
>> I want the master file to include a column with scores as shown below.
>> For example, ABC Inc gets a score of 7 in 1986, XYZ Corp gets a score
>> of 5 in 1994. For observations with two or more names, I want a simple
>> average of scores. For example, for ABC Inc XYZ Corp, the score will
>> be (9+6)/2=7.5. If a company has a year in the master file, then it
>> definitely has a score for that year (i.e. time period) in the second
>> name year score
>> ABC Inc 1986 7
>> XYZ Corp 1994 5
>> ABC Inc XYZ Corp 2001 7.5
>> PNG LLC 2005 7
>> XYZ Corp PNG LLC 2007 6.5
>> I am using Stata 12. I expect I would need the -regexm()- command to
>> split the company names, then -reshape- to get all company names one
>> below the other and then -merge-. Since there are 1100 companies, I
>> would need some kind of a loop to use the -regexm() command. I am
>> having trouble writing the code.
> * 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/
* For searches and help try: