How to use Propensity Matched Score method in MATLAB?

I have a large patient table with many data sets (rows) and variables (columns).

I have two groups of patients with group A (var1='ON') and group B (var1='off').

Now, I would like to create matching pairs of patients so that one patient in group A corresponds to one patient in group B dependent on a set of variables, e.g. age, gender, study days, etc..

Is there a Matlab method/function where I can easily create the second group B from a given group A and a given set of categorical and numerical variables with optimized matching? In literature, it's often called the propensity-matched score method.

Answer 1

[MathWorks Support Team]

It seems that there are two steps involved in Propensity Score Matching (PSM):

(1) Propensity score computation

(2) Some kind of observation matching based on the scores computed in (1).

Step (1) involves modeling the probability of receiving a treatment as a function of several predictors. Logistic regression seems to be a common way of doing step (1) and you can try to use 'glmfit' or 'glmval' for this step.

It seems that you are more interested in step (2) and as of now, we currently do not offer a function to do this. However, we already have an enhancement request to implement this feature in a future MATLAB release.

In addition to that, there is a File Exchange submission that claims to do PSM, which might be useful for you:

https://www.mathworks.com/matlabcentral/fileexchange/104730-bayesian-spatial-psm