How to generate dummy variable in stata having string multiple response

[Ashish Bandhu]

Dear All,

I want to generate dummy variables from a string variable cards_hh which has response like ABCD (without any space). I have tried split command but gives me error.


split cards_hh , generate(resp) destring
cannot generate new variables using stub resp
r(110);


. split cards_hh , generate(resp) destring
variable born as string:
resp1
resp1: contains nonnumeric characters; no replace




Any help is highly appreciated.


Thanks

Ashish

* Example generated by -dataex-. To install: ssc install dataex
clear
input str4 cards_hh
"B"   
"A"   
"AC"  
"C"   
"AC"  
"AC"  
"A"   
"A"   
"AB"  
"AB"  
"AC"  
"AC"  
"AC"  
"ABCD"
"ABC" 
"AC"  
"AC"  
"A"   
"AC"  
"A"   
"A"   
"A"   
"ACD" 
"ABC" 
"AC"  
"ABC" 
"D"   
"ABC" 
"AC"  
"AC"  
"AC"  
"AC"  
"AC"  
"AC"  
"ABC" 
"AB"  
"AC"  
"AC"  
"AC"  
"AC"  
"AC"  
"AC"  
"AC"  
"AC"  
"AC"  
"AC"  
"ACD" 
"ABD" 
"ACD" 
"ACD" 
"ABCD"
"ABD" 
"ABD" 
"ACD" 
"ACD" 
"ACD" 
"ACD" 
"ACD" 
"ACD" 
"ACD" 
"ACD" 
"ACD" 
"ACD" 
"ABC" 
"ACD" 
"AD"  
"ACD" 
"ACD" 
"D"   
"ACD" 
"ABC" 
"AC"  
"AC"  
"A"   
"AC"  
"ABD" 
"ACD" 
"AB"  
"ACD" 
"AC"  
"AC"  
"ABC" 
"AC"  
"ACD" 
"ABC" 
"AC"  
"C"   
"C"   
"AC"  
"C"   
"ABC" 
"C"   
"AC"  
"AC"  
"AC"  
"AC"  
"ABC" 
"AC"  
"A"   
"AC"  
end

[Carlo Lazzaro]

Ashish:
why not considering:

. egen wanted=group(cards_hh)
. label define wanted 1 "A" 2 "AB" 3 "ABC" 4 "ABCD" 5 "ABD" 6 "AC" 7 "ACD" 8 "AD" 9 "B" 10 "C" 11 "D"
. label val wanted wanted

[Nick Cox]

split requites a separator, such as a space. You don't have any separator.

As its original author I can testify that I considered extending its application to cases like this, but pulled back on two grounds:

1. That would complicate the syntax and the help file.

2. There is generally a simple alternative.

Point 2 above is true in this case.

In any case split doesn't generate indicator variables (you say "dummy") at all.

foreach value in A B C D { 
       gen `value' = strpos(cards_hh, "`value'") > 0 
}

will produce four indicator variables that are 1 if the named character occurs and 0 otherwise.

See https://www.stata-journal.com/articl...article=dm0099 Section 2 for a recommendation against the ugly term "dummy variable".

[Ashish Bandhu]

Ashish:
why not considering:

. egen wanted=group(cards_hh)
. label define wanted 1 "A" 2 "AB" 3 "ABC" 4 "ABCD" 5 "ABD" 6 "AC" 7 "ACD" 8 "AD" 9 "B" 10 "C" 11 "D"
. label val wanted wanted

Thank you for reply. I did not want to encode the original variable rather wanted generate 4 dummy variables based on response A, B, C, D. Cox has solved the problem.

[Ashish Bandhu]

split requites a separator, such as a space. You don't have any separator.

As its original author I can testify that I considered extending its application to cases like this, but pulled back on two grounds:

1. That would complicate the syntax and the help file.

2. There is generally a simple alternative.

Point 2 above is true in this case.

In any case split doesn't generate indicator variables (you say "dummy") at all.

foreach value in A B C D {
gen `value' = strpos(cards_hh, "`value'") > 0
}

will produce four indicator variables that are 1 if the named character occurs and 0 otherwise.

See https://www.stata-journal.com/articl...article=dm0099 Section 2 for a recommendation against the ugly term "dummy variable".

Thank you Nick. Got it