How to link Children's Stunting Status to Education Outcomes in Panel Data

[Charles Ackah]

Dear all,
I have a balanced panel of children observed for 3 years: 2012, 2014, and 2016. I have information on the stunting status (stunting) and whether or not the child successfully completed a school grade (grade_completed). The children are aged 0 to 14 years: range from 0-10 years in 2012 (wave 1), and 4-14 years in 2016 (wave 3). I want to establish a link between the children's stunting status in 2012, and their grade completion status in 2016. Both stunting and grade completion are binary variables with a value of 1 if the child is stunted or completed the school grade and 0 otherwise. Specifically, I want to know the distribution of grade completion in 2016 by the children's stunting status in 2012. I want to answer the following questions:

1. What proportion of children who were stunted in 2012 completed the school grade in 2016?
2. What proportion of children who were not stunted in 2012 completed the school grade in 2016?

I have tried different codes, but my results are not convincing, and I sometimes get no observation when I do a two-way tabulation: My example data is below:

* Example generated by -dataex-. To install: ssc install dataex
clear
input long pid float(wave age stunting grade_completed age_grp)
401014 3  7 0 0 1
401014 4  9 0 0 1
401014 5 11 0 0 2
401016 5 13 0 0 2
401016 3  8 0 0 1
401016 4 11 0 0 2
401018 5 12 0 0 2
401018 3  7 0 0 1
401018 4 10 0 0 1
401020 4  8 0 0 1
401020 5 10 0 0 1
401020 3  6 0 0 1
401023 5 11 0 0 2
401023 3  6 0 0 1
401023 4  9 0 0 1
401024 4  7 0 0 1
401024 5  9 0 0 1
401024 3  4 0 0 1
401028 5 13 0 0 2
401028 4 11 0 0 2
401028 3  9 0 0 1
401030 3  5 1 0 1
401030 5 10 0 0 1
401030 4  7 0 0 1
401032 4  8 1 0 1
401032 3  5 1 0 1
401032 5 10 1 0 1
401036 5 13 0 0 2
401036 3  9 0 0 1
401036 4 11 0 0 2
401037 4  9 0 0 1
401037 3  7 0 0 1
401037 5 12 0 0 2
401038 5 12 0 0 2
401038 4  9 0 0 1
401038 3  7 0 0 1
401040 4 11 0 0 2
401040 3  8 0 0 1
401040 5 13 0 0 2
401043 5 10 0 0 1
401043 4  7 0 0 1
401043 3  5 0 0 1
401044 5 14 0 0 2
401044 4 12 0 0 2
401044 3 10 1 0 1
401045 3  5 0 0 1
401045 5 10 0 0 1
401045 4  7 0 0 1
401048 5 11 0 0 2
401048 4  8 0 0 1
401048 3  6 0 0 1
401050 4 10 0 0 1
401050 3  8 0 0 1
401050 5 12 0 0 2
401052 5 13 0 0 2
401052 4 11 0 0 2
401052 3  9 0 0 1
401056 5 14 0 0 2
401056 3  9 0 0 1
401056 4 12 0 0 2
401064 4  8 0 0 1
401064 3  6 0 0 1
401064 5 11 0 0 2
401065 4 10 0 0 1
401065 3  7 0 0 1
401065 5 12 0 0 2
401066 3  8 1 0 1
401066 4 10 1 0 1
401066 5 13 1 0 2
401068 4  8 0 0 1
401068 3  5 0 0 1
401068 5 10 0 0 1
401070 5  9 0 1 1
401070 4  6 0 1 1
401070 3  4 0 0 1
401073 5 11 0 0 2
401073 4  9 0 0 1
401073 3  7 0 0 1
401084 3  8 0 0 1
401084 5 12 0 0 2
401084 4 10 0 0 1
401091 5 14 1 0 2
401091 4 11 1 0 2
401091 3  9 1 0 1
401095 3  8 1 0 1
401095 5 13 1 0 2
401095 4 11 1 0 2
401097 5 11 0 0 2
401097 4  9 0 0 1
401097 3  6 0 0 1
401106 5 13 0 0 2
401106 4 11 0 0 2
401106 3  8 0 0 1
401111 5 13 0 0 2
401111 3  9 0 0 1
401111 4 11 0 0 2
401112 5 11 0 0 2
401112 3  6 0 0 1
401112 4  8 0 0 1
401120 5 12 0 0 2
end
label values stunting stunting_lbl
label def stunting_lbl 0 "0. Not Stunted", modify
label def stunting_lbl 1 "1. Stunted", modify
label values grade_completed grade_completed
label def grade_completed 0 "No", modify
label def grade_completed 1 "Yes", modify
label values age_grp age_grp
label def age_grp 1 "0 to 10 years", modify
label def age_grp 2 "11 to 14 years", modify

[Clyde Schechter]

Well, you cannot do this with the data shown. You need information on stunting status in wave 1 for that, but your data starts with wave 3.

If you had wave 1 in the data, here's how you would do it:

isid pid wave, sort
by pid (wave): egen stunting_2012 = max(cond(wave == 1, stunting, .))
by pid (wave): egen grade_completed_2016 = max(cond(wave == 3, grade_completed, .))
egen tag = tag(pid)
tab grade_completed_2016 stunting_2012 if tag, col

[Charles Ackah]

Thanks Clyde Schechter . I have wave 1 information in my data.