Hello,
I believe this is a very newbie question but would be greatly appreciated if anyone can help me.


My test for covariates balancing using a code (tebalance overid) shows my two groups are not balanced.

I have been searching for another way to run PSM without violating the rule of covariates imbalance and just found that I may be able to run an analysis using a smaller block satisfying a condition that propensity score is not different for treated and controls.

Here is the command.

Based on the result below, the final number of block 8 seems good to run an analysis (the mean propensity score is not different for treated and controls).

I wonder what would be the next step and how to run it?
Also, if there are any materials addressing this topic, can you please share it with me?

Thanks in advance!!


Meenhye


pscore compare_adult_2 age_p sex poverty racreci3 new_maritl newwrklyr4, pscore(myscore) blockid (myblock)



************************************************** **
Algorithm to estimate the propensity score
************************************************** **


The treatment is compare_adult_2

Parent |
Comparison |
based on |
kids with |
ASD | Freq. Percent Cum.
------------+-----------------------------------
0 | 77,412 97.75 97.75
1 | 1,783 2.25 100.00
------------+-----------------------------------
Total | 79,195 100.00



Estimation of the propensity score

Iteration 0: log likelihood = -3496.7089
Iteration 1: log likelihood = -3439.9043
Iteration 2: log likelihood = -3438.7454
Iteration 3: log likelihood = -3438.7442

Probit regression Number of obs = 29731
LR chi2(6) = 115.93
Prob > chi2 = 0.0000
Log likelihood = -3438.7442 Pseudo R2 = 0.0166

------------------------------------------------------------------------------
compare_ad~2 | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
age_p | .007069 .0014693 4.81 0.000 .0041892 .0099488
sex | -.0121661 .0337414 -0.36 0.718 -.0782981 .0539659
poverty | -.0223613 .0100622 -2.22 0.026 -.0420829 -.0026397
racreci3 | -.140606 .0264132 -5.32 0.000 -.192375 -.088837
new_maritl | .1637391 .0343988 4.76 0.000 .0963188 .2311595
newwrklyr4 | -.1898755 .0361577 -5.25 0.000 -.2607433 -.1190078
_cons | -1.916608 .098875 -19.38 0.000 -2.110399 -1.722816
------------------------------------------------------------------------------



Description of the estimated propensity score

Estimated propensity score
-------------------------------------------------------------
Percentiles Smallest
1% .0080894 .0005976
5% .0118862 .0045809
10% .0154703 .0046339 Obs 42,571
25% .0187592 .004692 Sum of Wgt. 42,571

50% .023505 Mean .0261764
Largest Std. Dev. .011498
75% .0311225 .0939295
90% .0401195 .0959851 Variance .0001322
95% .0482016 .0959851 Skewness 1.593812
99% .0682563 .0959851 Kurtosis 7.047275



************************************************** ****
Step 1: Identification of the optimal number of blocks
Use option detail if you want more detailed output
************************************************** ****


The final number of blocks is 8

This number of blocks ensures that the mean propensity score
is not different for treated and controls in each blocks



************************************************** ********
Step 2: Test of balancing property of the propensity score
Use option detail if you want more detailed output
************************************************** ********


The balancing property is satisfied


This table shows the inferior bound, the number of treated
and the number of controls for each block

| Parent Comparison
Inferior | based on kids with
of block | ASD
of pscore | 0 1 | Total
-----------+----------------------+----------
0 | 50,218 1,050 | 51,268
.0125 | 5,524 96 | 5,620
.01875 | 5,967 111 | 6,078
.021875 | 3,841 109 | 3,950
.025 | 5,605 152 | 5,757
.03125 | 3,465 117 | 3,582
.0375 | 1,988 101 | 2,089
.05 | 804 47 | 851
-----------+----------------------+----------
Total | 77,412 1,783 | 79,195


*******************************************
End of the algorithm to estimate the pscore
*******************************************