Making Food Insecurity Index

[aafaque ali]

Using This dataset
. dataex

----------------------- copy starting from the next line -----------------------

* Example generated by -dataex-. For more info, type help dataex
clear
input long hhcode byte(province region) long psu int district byte(c01 c02 c03 c04 c05 c06 c07 c08)
101100101 1 1 1011001 101 2 2 2 2 2 2 2 2
101100102 1 1 1011001 101 2 2 2 2 2 2 2 2
101100103 1 1 1011001 101 2 2 2 2 2 2 2 2
101100104 1 1 1011001 101 1 1 2 2 2 2 2 2
101100105 1 1 1011001 101 1 1 1 2 1 2 2 2
101100106 1 1 1011001 101 2 1 1 2 1 2 2 2
101100107 1 1 1011001 101 2 2 2 2 2 2 2 2
101100108 1 1 1011001 101 2 1 2 2 2 2 2 2
101100109 1 1 1011001 101 2 2 2 2 2 2 2 2
101100110 1 1 1011001 101 2 2 2 2 2 2 2 2
101100111 1 1 1011001 101 2 2 2 2 2 2 2 2
101100112 1 1 1011001 101 2 1 2 2 2 2 2 2
101100113 1 1 1011001 101 1 1 1 2 1 2 2 2
101100114 1 1 1011001 101 1 1 1 2 1 1 2 2
101100115 1 1 1011001 101 2 2 2 2 1 2 2 2
101100116 1 1 1011001 101 1 1 1 2 2 2 1 1
101100117 1 1 1011001 101 2 2 2 2 2 2 2 2
101100118 1 1 1011001 101 1 1 1 2 2 2 2 2
101100119 1 1 1011001 101 1 2 1 2 2 1 2 2
101100120 1 1 1011001 101 1 1 2 2 2 2 2 2
101100121 1 1 1011001 101 2 2 2 2 2 2 2 2
101100122 1 1 1011001 101 1 1 2 2 2 2 2 2
101100123 1 1 1011001 101 2 2 2 2 2 2 2 2
101100124 1 1 1011001 101 1 2 1 2 2 2 2 2
101100125 1 1 1011001 101 2 2 2 2 2 2 2 2
101100126 1 1 1011001 101 1 1 1 2 2 2 2 2
101100127 1 1 1011001 101 2 2 2 2 2 2 2 2
101100128 1 1 1011001 101 1 1 1 2 2 2 2 2
101100129 1 1 1011001 101 2 2 2 2 2 2 2 2
101100130 1 1 1011001 101 1 2 2 2 2 2 2 2
101100201 1 1 1011002 101 2 2 2 2 2 2 2 2
101100202 1 1 1011002 101 2 2 2 2 2 2 2 2
101100203 1 1 1011002 101 2 2 2 2 2 2 2 2
101100204 1 1 1011002 101 2 2 2 2 2 2 2 2
101100205 1 1 1011002 101 2 2 2 2 2 2 2 2
101100206 1 1 1011002 101 1 1 1 2 1 2 2 2
101100207 1 1 1011002 101 2 2 2 2 2 2 2 2
101100208 1 1 1011002 101 2 2 2 2 2 2 2 2
101100209 1 1 1011002 101 2 2 2 2 2 2 2 2
101100210 1 1 1011002 101 2 1 1 2 2 2 2 2
101100211 1 1 1011002 101 2 2 2 2 2 2 2 2
101100212 1 1 1011002 101 2 2 2 2 2 2 2 2
101100213 1 1 1011002 101 2 2 2 2 2 2 2 2
101100214 1 1 1011002 101 2 2 2 2 2 2 2 2
101100215 1 1 1011002 101 2 2 2 2 2 2 2 2
101100217 1 1 1011002 101 2 2 2 2 2 2 2 2
101100218 1 1 1011002 101 1 1 1 2 2 2 1 1
101100219 1 1 1011002 101 2 2 1 2 2 2 2 2
101100220 1 1 1011002 101 2 2 2 2 2 2 2 2
101100221 1 1 1011002 101 2 2 2 2 2 2 2 2
101100222 1 1 1011002 101 2 2 1 2 2 2 2 2
101100223 1 1 1011002 101 2 2 1 2 2 2 2 2
101100224 1 1 1011002 101 2 2 2 2 2 2 2 2
101100225 1 1 1011002 101 2 2 2 2 2 2 2 2
101100226 1 1 1011002 101 2 2 1 2 2 2 2 2
101100227 1 1 1011002 101 2 2 2 2 2 2 2 2
101100228 1 1 1011002 101 2 2 1 2 2 2 2 2
101100229 1 1 1011002 101 2 2 2 2 2 2 2 2
101100230 1 1 1011002 101 1 1 1 2 2 2 2 2
101100401 1 1 1011004 101 2 2 2 2 2 2 2 2
101100402 1 1 1011004 101 2 2 2 2 2 2 2 2
101100403 1 1 1011004 101 2 2 2 2 2 2 2 2
101100404 1 1 1011004 101 2 2 2 2 2 2 2 2
101100405 1 1 1011004 101 2 2 2 2 2 2 2 2
101100406 1 1 1011004 101 2 2 2 2 2 2 2 2
101100407 1 1 1011004 101 2 2 2 2 2 2 2 2
101100408 1 1 1011004 101 2 2 2 2 2 2 2 2
101100409 1 1 1011004 101 1 1 2 2 2 2 2 2
101100410 1 1 1011004 101 2 2 2 2 2 2 2 2
101100412 1 1 1011004 101 2 2 2 2 2 2 2 2
101100413 1 1 1011004 101 1 2 1 2 2 2 2 2
101100414 1 1 1011004 101 2 2 2 2 2 2 2 2
101100415 1 1 1011004 101 2 2 2 2 2 2 2 2
101100416 1 1 1011004 101 2 2 2 2 2 2 2 2
101100417 1 1 1011004 101 2 2 2 2 2 2 2 2
101100418 1 1 1011004 101 2 2 2 2 2 2 2 2
101100420 1 1 1011004 101 2 2 2 2 2 2 2 2
101100421 1 1 1011004 101 2 2 2 2 2 2 2 2
101100422 1 1 1011004 101 2 2 1 2 2 2 2 2
101100423 1 1 1011004 101 2 2 2 2 2 2 2 2
101100424 1 1 1011004 101 2 2 2 2 2 2 2 2
101100425 1 1 1011004 101 2 2 2 2 2 2 2 2
101100426 1 1 1011004 101 1 1 1 2 2 2 2 2
101100427 1 1 1011004 101 2 2 2 2 2 2 2 2
101100428 1 1 1011004 101 2 2 2 2 2 2 2 2
101100429 1 1 1011004 101 2 2 2 2 2 2 2 2
101100430 1 1 1011004 101 2 2 2 2 2 2 2 2
101100501 1 1 1011005 101 2 2 2 2 2 2 2 2
101100502 1 1 1011005 101 2 2 2 2 2 2 2 2
101100503 1 1 1011005 101 2 2 2 2 2 2 2 2
101100504 1 1 1011005 101 2 2 2 2 2 2 2 2
101100505 1 1 1011005 101 1 1 1 2 1 2 2 2
101100506 1 1 1011005 101 1 1 1 2 1 2 2 2
101100507 1 1 1011005 101 1 1 1 2 2 2 2 2
101100508 1 1 1011005 101 2 2 2 2 2 2 2 2
101100509 1 1 1011005 101 1 1 1 2 2 2 2 2
101100510 1 1 1011005 101 2 2 2 2 2 2 2 2
101100511 1 1 1011005 101 2 2 1 2 2 2 2 2
101100512 1 1 1011005 101 2 2 1 2 2 2 2 2
101100513 1 1 1011005 101 2 2 1 2 2 2 2 2
end
label values province province
label def province 1 "kp", modify
label values region region
label def region 1 "rural", modify
label values district district
label def district 101 "abbottabad", modify
keep if province==2
drop if c01 == 98 | c01 == 99
drop if c02 == 98 | c02 == 99
drop if c03 == 98 | c03 == 99
drop if c04 == 98 | c04 == 99
drop if c05 == 98 | c05 == 99
drop if c06 == 98 | c06 == 99
drop if c07 == 98 | c07 == 99
drop if c08 == 98 | c08 == 99
recode c01 (1=1 "1") (2=0 "0"), generate(binary_c01)
recode c02 (1=1 "1") (2=0 "0"), generate(binary_c02)
recode c03 (1=1 "1") (2=0 "0"), generate(binary_c03)
recode c04 (1=1 "1") (2=0 "0"), generate(binary_c04)
recode c05 (1=1 "1") (2=0 "0"), generate(binary_c05)
recode c06 (1=1 "1") (2=0 "0"), generate(binary_c06)
recode c07 (1=1 "1") (2=0 "0"), generate(binary_c07)
recode c08 (1=1 "1") (2=0 "0"), generate(binary_c08)
collapse (mean) binary_c01 binary_c02 binary_c03 binary_c04 binary_c05 binary_c06 binary_c07 binary_c08 , by(district)
foreach var of varlist binary_c01 binary_c02 binary_c03 binary_c04 binary_c05 binary_c06 binary_c07 binary_c08 {
egen min_`var' = min(`var')
egen max_`var' = max(`var')
gen normalized_`var' = (`var' - min_`var') / (max_`var' - min_`var')
}
list binary_c01 normalized_binary_c01 in 1/10
gen food_insecurity_index = (normalized_binary_c01 + normalized_binary_c02 + normalized_binary_c03 + normalized_binary_c04 + normalized_binary_c05 + normalized_binary_c06 + normalized_binary_c07 + normalized_binary_c08) / 8
list district food_insecurity_index
------------------ copy up to and including the previous line ------------------

Listed 100 out of 160654 observations
Use the count() option to list more
Result:
district food_i~x

1. attock .0878222
2. bahawaln .7673687
3. bahawalp .6395296
4. bhakhar .384757
5. chakwal .2943359

6. chiniot .2689224
7. d. g. kh .5084416
8. faisalab .1328354
9. gujranwa .2444148
10. gujrat .0180971

11. hafizaba .2025424
12. islamaba .1586573
13. jehlum .4989371
14. jhang .5090191
15. kasur .7063363

16. khanewal .7514324
17. khushab .2625226
18. lahore .4137256
19. layyah .4822337
20. lodhran .6053461

21. mandi ba .1896518
22. mianwali .2228872
23. multan .3976084
24. muzaffar .6404075
25. nankana .2188804

26. narowal .3084604
27. okara .0613527
28. pakpatta .2036972
29. rahim ya .4872369
30. rajanpur .3854529

31. rawalpin .1842262
32. sahiwal .1838323
33. sargodha .3961037
34. sheikhup .3443179
35. sialkot .1976983

36. t.t. sin .2923453
37. vehari .3706231

I just want to ask whether this is the correct method for making a food insecurity index?
Thank You for Your Assistance

[Daniel Schaefer]

This is the relevant portion of code from what I can tell.

foreach var of varlist binary_c01 binary_c02 binary_c03 binary_c04 binary_c05 binary_c06 binary_c07 binary_c08 {
    egen min_`var' = min(`var')
    egen max_`var' = max(`var')
    gen normalized_`var' = (`var' - min_`var') / (max_`var' - min_`var')
}
list binary_c01 normalized_binary_c01 in 1/10
gen food_insecurity_index = (normalized_binary_c01 + normalized_binary_c02 + normalized_binary_c03 + normalized_binary_c04 + normalized_binary_c05 + normalized_binary_c06 + normalized_binary_c07 + normalized_binary_c08) / 8

The for loop is a little strange since it looks like you are treating the data as quantitative when each variable is in fact dichotomous. The minimum for each var should be zero every time, so when you subtract off min_`var' that isn't having any effect. This leaves you "normalizing" by dividing each binary variable by one, so if 1, it should be 1, and if 0 it should remain zero. There is no reason I can see to do this. Since each variable is binary, I actually don't think you need to normalize here at all.

The last line should be okay. Whether or not you exclude the for loop and don't try to normalize your dichotomous variables, the result of the last line is interpretable as the proportion of the binary variables with a 1. Notice that if you want this to be an index, the meaning of the ones have to go in the same direction. So if 1 means the presence of some feature of food insecurity on one variable, it has to mean the presence (not absense) of some feature of food insecurity on every other variable.

Edit: I assume you used a guide to finish this? Just to be clear, what your for loop does is it takes a quantitative value and rescales it to be between 0 and 1. Since dichotomies variables are already on a 0/1 scale, you don't need to rescale.

[Daniel Schaefer]

By the way, "correct" is a relative term. There are better methods in confirmatory factor analysis. Different fields construct indexes differently, so the "best" way to do it might be field dependent.