Creating Binary variable

[Shamsudini Amidu]

Hi Family,
I have a household data set containing nutritional outcome of at least 2 children in each household.
I am seeking to create an additional variable named "divergent_nut_outcome" with a binary outcome, where 1 signifies 'Divergent' (i.e., if children within the same household exhibit differing nutritional outcomes), and 0 indicates 'Not Divergent' (all children within the household have the same nutritional status, either undernutrition or normal).

This is the command I tried but became confused along the line.

egen underNutriton_count = total( nutritional_status == 2), by( nutritional_status )
gen divergent_status = "Divergent" if underNutriton_count > 0 & underNutriton_count < _N

I Kindly request your assistance to resolve this.

Below is the example of the data set for your perusal.
[CODE]
* Example generated by -dataex-. For more info, type help dataex
clear
input str3 uid byte age_months long(gender nutritional_status)
"119" 52 2 1
"119" 21 1 2
"119" 5 1 1
"131" 40 1 1
"131" 5 2 2
"311" 54 2 1
"311" 11 2 2
"419" 39 1 1
"419" 59 2 1
"419" 19 1 1
"525" 33 2 2
"525" 21 2 1
"526" 57 1 1
"526" 8 1 1
"633" 39 1 1
"633" 34 1 1
"7 2" 27 2 2
"7 2" 7 2 2
end
label values gender gender
label def gender 1 "female", modify
label def gender 2 "male", modify
label values nutritional_status nutritional_status
label def nutritional_status 1 "Normal", modify
label def nutritional_status 2 "Undernutrition", modify

Thank you for your continues support.
Amidu

[Ken Chui]

There is no variation in the status codes if all children have the same outcome within a household, that means the HH-level SD will be 0. So, one way to solve this would be through egen & sd():

egen divergent_nut_outcome = sd(nutritional_status), by(uid)
replace divergent_nut_outcome = 1 if divergent_nut_outcome > 0 & !missing(divergent_nut_outcome)

[Shamsudini Amidu]

Hi ken,
I am most grateful for your swift response.
It has created it successfully, I am running a logistics regression to investigate whether the gender of the child is associated with the nutritional divergent.

I use the command below;
logistic divergent_nut_outcome i.gender


[CODE]
* Example generated by -dataex-. For more info, type help dataex
clear
input str3 uid byte age_months long(gender nutritional_status) float divergent_nut_outcome
"119" 52 2 1 1
"119" 21 1 2 1
"119" 5 1 1 1
"131" 40 1 1 1
"131" 5 2 2 1
"311" 54 2 1 1
"311" 11 2 2 1
"419" 39 1 1 0
"419" 59 2 1 0
"419" 19 1 1 0
"525" 33 2 2 1
"525" 21 2 1 1
"526" 57 1 1 0
"526" 8 1 1 0
"633" 39 1 1 0
"633" 34 1 1 0
"7 2" 27 2 2 0
"7 2" 7 2 2 0
end
label values gender gender
label def gender 1 "female", modify
label def gender 2 "male", modify
label values nutritional_status nutritional_status
label def nutritional_status 1 "Normal", modify
label def nutritional_status 2 "Undernutrition", modify


Am I correct?

Alternatively, I also taught of collapsing the divergent_nut_outcome by household' before I run the logistic regression,
However, if I do so, I will face with potential bias when collapsing the data to a single observation per household and the risk of losing individual-level information like sex, age.

What is the best way to go?

Thank you.
I do appreciate your time and support.