splitting a string variable into unique dummy variables

[Jean-Luc Morin-Chesnel]

Dear all,

I have a string-related problem that I was not able to solve so far. I use Stata 13 MP.

I have string variable "product_code" that, at each date, contains product codes separated by blank spaces

1	STATA SAS MATLAB
2	R MATLAB
3	STATA EXCEL PEN

I would like to create dummy variables for each possible product code in my dataset so that I can know at each date which product code was available.
In the example, the dataset I want to obtain looks like

date	product code	D_STATA	D_SAS	D_MATLAB	D_R	D_EXCEL	D_PEN
1	STATA SAS MATLAB	1	1	1	0	0	0
2	R MATLAB	0	0	1	1	0	0
3	STATA EXCEL PEN	1	0	0	0	1	1

Do you know how I can do that? I tried using -split, but without success.
Many thanks!!

[Nick Cox]

In this case,

 
foreach s in Stata SAS MATLAB R Excel pen { 
     gen D_`s' = strpos(product_code, "`s'") > 0 
}

In practice, watch for lower and upper case and spelling differences.

[Jean-Luc Morin-Chesnel]

Thank you Nick!!
Neat implementation indeed.

Assume however that the list of possible product codes is just too large to be typed manually in the loop.
I was thinking about using levelsof to get the unique product codes in the dataset, but the fact that multiple product codes are contained in one single cell complicates the matter...

Do you see any possible solution here?
Thanks again

[Nick Cox]

Uli Kohler and I wrote a paper on this some years ago. http://www.stata-journal.com/sjpdf.h...iclenum=pr0008 I don't want to re-read it to rediscover if we discussed this.

But you have all the ingredients of an answer already:

 
split product_code 

foreach v in `r(varlist)' {
     levelsof `v', local(this)
     local all : list all | this
}

mac li

If your items are not single words, it's a lot messier.

[Jean-Luc Morin-Chesnel]

Of course, everything is stored in r(varlist) !!!
Thank you very much again, Nick