Balance of covariates within blocks of Propensity scores

[Denise Vella]

Hi I've been working and using dummy data to learn Propensity scores - and this article has really helped me
https://pubmed.ncbi.nlm.nih.gov/24779867/

I'm not sure if people David Radwin on the forum can answer my questions, or perhaps I should email Prof Melissa Garrido directly.


Here are my questions;

I have used this data:

webuse cattaneo2

Research question:
Does alochol effect birth weight in ?

Matching on: maternal age, maternal education, maternal smoking.

**Creating macros
global treatment alcohol //intervention/treatment
global ylist bweight //outcome
global xlist mage medu msmoke //variables to match on t

QUESTION 1
Prof Garrido says:
Common support is subjectively assessed by examining a graph of propensity scores across treatment and comparison groups/ Besides overlapping - balance in treated and comparison groups is splitting the sample to ensure the mean propensity score distribution can be obtained.

Now, I know propensity scores can be calculated using -psmatch2 - and

pscore $treatment $xlist,pscore(myscore)blockid(myblock)comsup detail

- How can you compare this mean in treated and comparison groups by using psmatch2? When the output doesn't show the mean differences. (SEE pics attached)
In the -pscore- all blocks are plotted and if there is a difference between the treatment and control the blocks are split until there is no difference.

QUESTION 2 - Main question
Prof Garrido writes - Step 3: Balancing co variates across the treatment and comparison groups with the blocks of propensity scores.
Variables should have a standarized difference <25% of the maximum SD.' But Prof Garrido does not show the data here.

Question 2A
I assume to get the maximum SD this would be??
summarize pscore //calculate the SD of the propensity score



QUESTION 2B - my main question
//However regarding phrase 'check for the balance of individual covariates across treatment and comparison groups'

Does this mean repeating the code but varying the macro xlist with different covariance to try get balance immediately and avoid splitting the blocks.
pscore $treatment $xlist,pscore(myscore)blockid(myblock)comsup detail

How do you do this?

[Denise Vella]

With reference to question 2:

I found

-pstest-

Not sure if this is correct

Does this mean that if %bias is <25 than that means there is balance between ALL blocks in treatment and control matched propensity score groups? (or perhaps <0.25

As -pstest- help says:

The standardised % bias is the % difference of the sample means in the treated and non-treated (full or matched)
sub-samples as a percentage of the square root of the average of the sample variances in the treated and non-treated groups (formulae from Rosenbaum and Rubin,
1985).

when Prof Garrido et al. cites ' proposed maximum standardized differences for specific co variates range from 10 to 25 per cent' So should I be squaring the value as a square root is presented in the table of results?

[Denise Vella]

George Ford can you help?