Propensity Score Matching to measure impact of an intervention

[Richard Edro]

I wanted to estimate impact of the use of indoor residual spraying on malaria morbidities and mortalities among children the under age of 5 years. I used the following codes;

logit IRS hw1 b4 hc61 v190 v113 v116 v216 h22 v228 v025
predict pscore
psmatch2 IRS Malaria, kernel common
teffects psmatch (Malaria) (IRS, pscore(pscore))

However code 4 gave me the following error;

. teffects psmatch (Malaria) (IRS, pscore(pscore))
option pscore() not allowed
    The treatment model is misspecified.
r(198);

Kindly help me to get the codes right and succeed in estimating the impact of the IRS program. below is the dataset upto the 100th observation

* Example generated by -dataex-. For more info, type help dataex
clear
input float(IRS Malaria) byte(hw1 b4 hc61 v190 v113 v116 v216 h22) float use_net byte v025
1 . 16 1 2 4 72 21 2 1 1 1
1 0 14 1 4 2 21 23 4 1 0 1
1 . 24 2 0 5 11 12 1 1 0 1
0 1 28 1 4 4 12 12 4 0 0 1
0 .  9 2 4 3 12 23 1 0 1 1
0 0 28 2 0 3 12 21 2 1 1 1
1 1  0 2 3 3 12 21 3 0 0 1
1 0 52 1 1 3 14 23 2 0 1 1
0 . 52 1 1 3 14 23 1 0 1 1
1 .  7 2 0 2 14 23 2 1 1 1
1 0 30 1 1 1 21 23 2 0 0 1
0 . 47 1 2 5 12 12 1 0 0 1
1 0 54 2 0 4 11 12 5 0 0 1
1 . 36 2 4 3 13 21 2 1 0 1
0 0 36 2 4 3 13 21 5 1 0 1
1 . 50 1 4 3 13 21 4 0 1 1
0 .  6 1 2 3 13 21 1 1 1 1
0 . 11 2 4 4 13 21 2 0 1 1
1 1 45 2 4 4 11 12 1 0 1 1
1 . 50 2 2 5 12 12 2 0 1 1
0 0  7 2 4 5 12 12 5 0 1 1
1 1  5 2 3 3 13 21 4 0 1 1
1 . 15 1 1 5 72 12 4 0 1 1
0 0 50 1 1 4 14 23 1 0 1 1
0 . 36 2 4 3 14 23 4 1 1 1
0 0  0 2 3 3 14 23 2 0 1 1
1 1 54 1 4 5 11 21 3 0 1 1
1 0  3 2 1 5 11 21 4 0 0 1
1 0 31 1 1 4 11 21 1 0 1 1
0 .  7 1 4 4 11 21 4 0 0 1
0 1 47 2 2 3 13 21 2 0 1 1
0 .  3 2 4 3 13 21 1 0 1 1
0 . 36 2 0 3 13 23 5 0 1 1
0 . 58 1 1 1 13 23 5 1 1 1
0 0  8 2 3 3 21 23 3 1 0 1
1 . 17 2 3 2 21 21 4 0 0 1
0 . 25 2 4 5 11 12 1 0 0 1
1 0  1 2 3 3 21 21 4 0 1 1
1 . 10 2 3 3 21 21 4 0 0 1
0 0 25 2 1 1 21 23 3 0 0 1
1 .  7 2 4 3 21 21 2 0 0 1
0 .  5 1 3 2 21 31 3 0 1 1
1 . 44 1 0 5 21 21 5 0 1 1
0 . 14 2 3 5 21 21 1 0 0 1
1 0 11 2 0 1 21 31 3 0 0 1
0 1 28 1 1 1 21 31 4 0 0 1
0 .  0 2 0 3 14 21 1 0 1 1
0 . 53 1 0 3 14 21 3 0 0 1
1 . 51 2 4 4 11 22 2 0 0 1
0 . 21 1 2 3 14 21 1 0 1 1
0 0 46 1 3 1 21 23 1 1 1 2
0 1 46 1 3 1 21 23 1 1 1 2
0 . 25 1 1 1 21 31 4 1 0 2
1 0 45 2 4 1 21 31 2 1 1 2
0 . 50 2 1 1 21 31 4 0 0 2
0 .  3 2 3 1 21 31 4 1 1 2
1 . 45 2 1 1 21 23 5 1 1 2
1 . 42 2 2 1 14 23 2 0 1 2
0 . 16 1 2 1 14 23 2 0 0 2
0 0 43 1 3 1 21 31 5 0 0 2
0 . 19 1 1 1 21 31 4 1 0 2
1 . 27 2 3 1 21 31 1 1 1 2
0 . 41 1 3 1 21 31 5 1 1 2
1 . 18 1 3 1 21 31 5 1 0 2
1 . 37 2 3 1 21 22 2 1 0 2
0 .  7 2 0 1 21 22 1 1 0 2
0 0 43 2 0 1 97 97 1 1 1 2
0 .  7 2 2 1 97 97 2 1 1 2
0 0  2 2 4 1 21 23 1 0 1 2
1 . 47 2 0 1 21 31 2 0 0 2
0 1 26 2 3 1 21 31 4 1 0 2
0 1 12 1 1 1 14 23 1 1 0 2
0 1  9 1 0 1 14 31 2 1 0 2
1 1 36 1 4 1 14 31 5 0 1 2
0 .  0 1 4 1 14 31 1 0 0 2
0 0  9 2 0 1 21 23 3 0 1 2
1 .  0 1 2 1 21 23 3 0 1 2
1 0  6 2 4 1 14 23 1 1 1 2
0 . 57 2 4 1 14 23 1 0 1 2
0 .  2 1 1 1 14 23 1 0 0 2
0 1 47 2 3 1 14 31 1 1 1 2
1 1  5 1 1 1 14 31 1 0 1 2
1 . 36 1 0 1 21 31 3 0 1 2
1 0 36 1 1 2 97 97 2 0 0 2
0 . 36 2 1 1 12 23 4 0 1 2
1 . 45 1 4 1 12 23 3 0 1 2
1 1  4 2 2 1 12 23 4 0 0 2
0 . 56 2 4 1 21 23 1 1 1 2
1 . 18 1 4 1 21 23 3 0 1 2
1 . 16 1 1 1 21 31 2 0 1 2
1 . 50 2 3 1 14 31 1 0 1 2
0 . 10 2 1 1 14 31 5 0 0 2
1 . 35 2 2 1 14 31 5 1 1 2
1 . 23 2 1 1 14 31 1 1 1 2
0 . 57 2 1 1 14 31 2 1 1 2
1 .  7 2 2 1 14 31 2 1 0 2
0 . 43 1 1 1 14 31 2 0 0 2
0 0  9 2 2 1 14 31 1 0 0 2
0 . 34 1 1 1 14 31 1 1 1 2
1 . 24 2 2 1 21 31 2 0 1 2
end
label values IRS yesno
label def yesno 0 "no", modify
label def yesno 1 "yes", modify
label values Malaria mal_status
label def mal_status 0 "negative", modify
label def mal_status 1 "positive", modify
label values b4 B4
label def B4 1 "male", modify
label def B4 2 "female", modify
label values hc61 HC61_01
label def HC61_01 0 "no education", modify
label def HC61_01 1 "primary", modify
label def HC61_01 2 "middle/jss/jhs", modify
label def HC61_01 3 "secondary/sss/shs", modify
label def HC61_01 4 "higher", modify
label values v190 V190
label def V190 1 "poorest", modify
label def V190 2 "poorer", modify
label def V190 3 "middle", modify
label def V190 4 "richer", modify
label def V190 5 "richest", modify
label values v113 V113
label def V113 11 "piped into dwelling", modify
label def V113 12 "piped to yard/plot", modify
label def V113 13 "piped to neighbor", modify
label def V113 14 "public tap/standpipe", modify
label def V113 21 "tube well or borehole", modify
label def V113 72 "sachet water", modify
label def V113 97 "not a dejure resident", modify
label values v116 V116
label def V116 12 "flush to septic tank", modify
label def V116 21 "ventilated improved pit latrine (vip)", modify
label def V116 22 "pit latrine with slab", modify
label def V116 23 "pit latrine without slab/open pit", modify
label def V116 31 "no facility/bush/field", modify
label def V116 97 "not a dejure resident", modify
label values h22 H22
label def H22 0 "no", modify
label def H22 1 "yes", modify
label values use_net yesno1
label def yesno1 0 "no", modify
label def yesno1 1 "yes", modify
label values v025 V025
label def V025 1 "urban", modify
label def V025 2 "rural", modify

[Felix Bittmann]

You do not need to generate the PS before with the ado, just type

teffects psmatch (Malaria) (IRS hw1 b4 hc61 v190 v113 v116 v216 h22 v025)

An Alternative is kmatch

ssc install kmatch, replace
kmatch ps IRS hw1 b4 hc61 v190 v113 v116 v216 h22 v025 (Malaria)

[Richard Edro]

Thanks Felix. Below is the result;

. teffects psmatch (Malaria) (IRS hw1 b4 hc61 v190 v113 v116 v216 h22 v025)

Treatment-effects estimation                   Number of obs      =        175
Estimator      : propensity-score matching     Matches: requested =          1
Outcome model  : matching                                     min =          1
Treatment model: logit                                        max =          1
------------------------------------------------------------------------------
             |              AI Robust
     Malaria |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
ATE          |
         IRS |
(yes vs no)  |   .1257143   .2392578     0.53   0.599    -.3432225     .594651
------------------------------------------------------------------------------

However, the expectation is show that IRS has an impact on Malaria but the results show that the ATE of IRS on Malaria is 0.1257143, which means that IRS increases the probability of having malaria by 12.57 percentage points on average. However, this effect is not statistically significant, as the p-value is 0.599, which is greater than the conventional significance level of 0.05. The 95% confidence interval for the ATE is [-0.3432225, 0.594651], which means that we are 95% confident that the true ATE of IRS on Malaria lies within this range. As it stands, we cannot conclude that IRS has a causal impact on malaria prevalence among children under the age of 5 years. Is there another way I could do this to prove there is an impact or it is okay to conclude this way. Already because this is limited to just region, the sample size small which could be a contributing factor