*IFA Intentions* March 15, 2021

log using "/Users/ericasedlander/Box/RANI Project/RCT/Baseline/Imputed/Paper Three.log", replace

use "/Users/ericasedlander/Box/RANI Project/RCT/Baseline/Imputed/PreImputation.dta", clear
 
set matsize 1000

*********************************************************************************************************************

drop if preg_stat == 1
*dropping all pregnant women 

*****HOUSEKEEPING *********

gen breast_feed = preg5
gen ifa_num = ifa2
*unevenly distributed 
gen current_ifa = ifa1
replace current_ifa = 0 if current_ifa == 2
label define current_ifalbl 0 "not taking ifa" 1 "taking ifa" 
label values current_ifa current_ifalbl
gen marital = preg2
gen religion = dem3
gen caste_gen = dem4
gen caste = dem5
gen child = dem6
gen phone = dem7
drop if educ == 99
*no response
gen hemo = anhaemwra
gen ever_anemia = dem17
gen current_anemia = dem18


*KNOWLEDGE* recoding knowledge questions so that 1=correct and 0=incorrect.
gen know1_new = 1 if know1 == 2 
replace know1_new = 0 if know1 == 1
replace know1_new = 0 if know1 == 3
tab know1
tab know1 know1_new 
tab know1_new

gen know2_new = 1 if know2 == 2 
replace know2_new = 0 if know2 == 1
replace know2_new = 0 if know2 == 3

gen know3_new = 1 if know3 == 1 
replace know3_new = 0 if know3 == 2
replace know3_new = 0 if know3 == 3

gen know4_new = 1 if know4 == 1 
replace know4_new = 0 if know4 == 2
replace know4_new = 0 if know4 == 3

gen know5_new = 1 if know5 == 1 
replace know5_new = 0 if know5 == 2
replace know5_new = 0 if know5 == 3

gen know6_new = 0 if know6 == 1
replace know6_new = 1 if know6 == 2
replace know6_new = 0 if know6 == 3

gen know7_new = 0 if know7 == 1
replace know7_new = 1 if know7 == 2
replace know7_new = 0 if know7 == 3


*CREATING INDEXES*
gen know = (know1_new + know2_new + know3_new + know4_new + know5_new + know6_new + know7_new)/7


*INTENTIONS*
gen intent = (int2+int3+int4)/3

alpha int2 int3 int4, item

polychoric int2 int3 int4
matrix A1 = r(R)
matrix B1 = diag(vec(A1))
local C1 = (trace(B1)-trace(A1))/2
local D1 = rowsof(A1)
local E1 = `C1'/(`D1'*(`D1'-1)/2)
local F1 = (`D1'*`E1')/(1+((`D1'-1)*`E1'))
display `F1'

*Descriptive IFA norms - non pregnant* did not include adolescents sn2 so only one item 
gen dsnorms = sn3
*what percentage of non pregnant women take IFA regularly. 77% said none and 20% said some. Not a lot of variance. 

*injunctive social norms - non preg women
gen isnorms = (sn5 + sn7+ sn9)/3

*OUTCOME EXPECTATIONS*
gen outcome_exp = oe2

*SELF EFFICACY*
gen self_eff = (se1+ se2 + se3 + se4)/4


*PERCEIVED IFA ACCESS*
gen per_access = pa2

*PERCEIVED RISK* *higher scores = higher percieved anemia risk 

gen risk = (rp1 + rp2)/2

*DATA EXPLORATION* - examples - ran histogram on every variable in the model to examine distribution, mean, and SD 

histogram dsnorms
histogram isnorms
histogram intent

*NON-SELF MEAN*

sort village
* sorted the dataset by village 
by village: egen  numberinvillage =count(village)
*numberinvillage = the # of people in each village
sum numberinvillage, d

**CREATING NON SELF MEANS** 

*intentions to take IFA non self mean
by village: egen intent_avg = mean(intent)
*created a new variable "intent_avg" which is the average intentions to take IFA for non preg women in each village 
by village: egen intent_sum = sum(intent)
by village: egen intent_count = count(intent)
*this counts the number of all people in each village
gen nonself_intent_mean = (intent_sum - intent) / (intent_count-1) 
gen intent_nsm = nonself_intent_mean
*this is reporting the nsm of intent 

*STANDARDIZING ALL VARIABLES THAT I MAY USE IN A MODEL THAT MAY BE DIFFICULT TO INTERPRET*

egen zdsnorms=std (dsnorms)
egen zisnorms=std (isnorms)
egen zeduc=std (educ)
egen zcaste=std (caste)
egen zknow=std (know) 
egen zmarital=std (marital)
egen zcurrent_ifa=std (current_ifa)
egen zintent=std (intent)
egen zoutcome_exp=std (outcome_exp)
egen zper_access=std (per_access) 
egen zrisk=std (risk) 
egen zself_eff=std (self_eff)
egen zbreast_feed=std (breast_feed)
egen zcurrent_anemia=std (current_anemia) 
egen zintent_nsm=std(intent_nsm)

*PEARSON CORRELATION TABLE*
pwcorr zintent zdsnorms zisnorms zknow zself_eff zrisk zper_access zoutcome_exp zintent_nsm, sig star (.05)
*All variables are significantly postively associated with IFA intentions 

**PEARSON CORRELATIONS*
pwcorr zdsnorms zintent, sig star (.05) 
pwcorr zisnorms zintent, sig star (.05) 
pwcorr current_ifa intent, sig star (.001)

*REGRESSIONS* 
regress zintent village zbreast_feed zmarital zcaste age child phone zself_eff zper_access zrisk zoutcome_exp zknow, vce (cluster village) 

vif
*Multicollinearity is NOT and issue - mean VIF = 1.22

 
*REGRESSIONS NON-PREGNANT WOMEN* *NESTED*
regress zintent village age zeduc child zbreast_feed zdsnorms zisnorms zintent_nsm, vce (cluster village) 
* model 1

regress zintent village age zeduc child zbreast_feed zdsnorms zisnorms zintent_nsm zknow zself_eff zrisk zper_access zoutcome_exp, vce (cluster village)
*model 2