Is it possible to destring a variable which starts with a letter?

[Ellie Daglish]

Hello, I am using Parliamentary Constituency codes in my dataset which begin with either E, S, N, W dependent on if the constituency is in England, Scotland, Northern Ireland or Wales respectively. The variable type is a string. I have made several attempts to destring the variable in the usual way however have been unsuccessful. I suspect this is because of the letter at the start of each observation? I was wondering whether there are possibly any suggestions as to overcome this? Many thanks.

[Iuri Gavronski]

Could you provide some examples of such variable?

[Ellie Daglish]

Here are some examples of the variables I am using:

E14000576
E14000876
S14000029
W07000687
W07000087
N06000045
N06000009

I want to drop all observations where the code begins with N however when I have attempted to do so it provides an error relating to the fact that they are in string format. Hence, my #1 question where I attempted to destring them so that I can drop these variables. If you have any suggestions of an alternative way to do this they would be greatly appreciated. Many thanks.

[Andrea Discacciati]

clear
input str9 code
"E14000576"
"E14000876"
"S14000029"
"W07000687"
"W07000087"
"N06000045"
"N06000009"
end

drop if substr(code, 1, 1) == "N"

Please, next time use -dataex- to post your example (see FAQ #12).

[Iuri Gavronski]

I would go

egen countryc=substr(code,1,1)
drop if countryc == "N"

HTH

[Ellie Daglish]

Thank you for your help, that is great.

[Nick Cox]

I guess people worked this out but in #5

 
 egen countryc=substr(code,1,1)

is a typo for

 
 gen countryc=substr(code,1,1)

Indeed, as Andrea Discacciati pointed out in #4 you don't need to create a new variable at all, as

drop if substr(code, 1, 1) == "N" 

gets you there directly.