Condition based on part of a string

[Ben Ronen]

Hi,
I want to execute some commands only on those observations that contain a specific phrase as part of their value in a string variable. What should be the condition?
P.S. the phrase isn't in English (Hebrew letters).
Thanks ahead,
Ben

[Igor Paploski]

Hi Ben,

The condition is up to you to come up with. One way you could implement this would be along the lines of:

gen test = 1 if strpos("string_to_look_for",var_you_are_looking_for_string)>0

What this does is it applies whatever you are doing (in this case, generating test=1) only to those observations in which the value for variable var_you_are_looking_for_string contains the string string_to_look_for. I'm not sure how Stata deals with Hebrew characters though.

Cheers

[Clyde Schechter]

I imagine that Hebrew characters are implemented in unicode, and so the condition proposed in #2 would use the -ustrpos()- function instead of -strpos()-.

[Ben Ronen]

Thank you, Igor and Clyde.

Actually, none of them works - it only get me those observations with an empty value in the specified variable (regardless the variable I specify and the language it holds, as I got the same results with English strings).

Do you have any idea what could be the cause?

[Igor Paploski]

Hey Ben, help strpos shows explains that the order I posted the suggestion to you is inverted.

Instead of:

gen test = 1 if strpos("string_to_look_for",var_you_are_looking_for_string)>0

It should be:

gen test = 1 if strpos(var_you_are_looking_for_string, "string_to_look_for")>0

This way, if you run the following:

sysuse auto.dta
gen buick = 1 if strpos(make,"Buick")>0

You will create a variable that is 1 for all entries that contain the word "Buick" (Stata is case-sensitive, "buick" would not be detected) in the variable make.