propensity score IPTW

[#1]

I am a jmp pro18 user.
I would like to perform propensity score matching using the propensity score inverse probability weighting (IPTW) method.
I would like to find an IPTW add-in, but I was unable to find one. What should I do? I would be grateful if you could tell me.

[#2]

Hi @keiichi, I'm not very familiar this IPTW, but would this link help? On it, there's a reply from @Chris_Kirchberg that has a link to a PDF (in Japanese). Page 8 of that PDF has a link to an Add-In that might be what you are looking for.

[#3]

Mr. Jed Campbell, thank you for your prompt reply. I know and use the PDF (in Japanese) of the reply from @Chris Kirchberg that you told me about. It is a very useful add-in for Propensity score matching in JMP. But you came across this PDF in Japanese. Thank you very much.

What I am looking to get this time is the Propensity score IPTW add-in. After searching, I found the following link.

https://community.jmp.com/t5/Abstracts/JMP-based inverse probability weighting using propensity scores and automation of execution procedures-Nakazono-Takasuke-2021-JA-25MP-12/ev-p/756540

The above link contains an article by Dr. Nakazono Takasuke from Discovery Summit Japan 2021 titled "JMP-based inverse probability weighting using propensity scores and automation of execution procedures." In the presentation, there is an explanation of an add-in that allows the IPTW method to be performed in JMP, but as I am not familiar with statistics, I do not fully understand it. The presenter, Dr. Nakazono Kosuke, appears to have already changed affiliation, and I have been unable to contact him. If I can understand the method of creating an IPTW add-in described in the presentation, I believe I will be able to create one in my own JMP. If anyone knows more about this, please let me know.

That's all, and I would be grateful for any guidance you can give me.

[#4]

I am afraid that "IPTW" and "propensity score matching" are different methods.

---
If in your IPTW analysis, you need to calculate only point estimates, and you do NOT need to calculate standard errors, p-values of statistical tests, and confidence intervals, then you can do IPTW (inverse probability of treatment weighting) by the following steps.

(1) Fit Logistic models for the binary treatment variable. Save the predicted probabilities by [Save Probability Formula].
(2) Calculate the inverse of probability by the formula (like by "(:Treatment==1)/:Prob1 + (:Treatment==0)/:Prob0".
(3) Fit a IPTW model with Y=your response variable, X=your treatment variable and Weight=IPTW calculated by (2).

Note that the standard errors, confidence intervals and p-values calculated by the above steps are statistically incorrect. This is because JMP's almost all platform assume weights as "precision weights", not "sampling weights".

In JMP Pro 19 (the next version of JMP Pro) or later, we will have a new platform, the Causal Treatment platform. So, you will can do IPTW by this platform (and standard errors, p-values and confidence intervals are also calculated correclty in this JMP Pro 19's new platform.)

In JMP Pro 18(the current version of JMP Pro) or before, there is no feature to calculate standard erorrs, p-values and confidence intervals in IPTW.

---
The propensity score matching will also be supported in JMP Pro 19(next version of JMP Pro).

Although it is just "as-is", I made an add-in to perform the one-to-one propensity score greedy matching about 10 years ago. The URL for downloading is written in the follwoing page.
https://www.jmp.com/content/dam/jmp/documents/jp/support/propensityscore201412.pdf

I would like to recommend that you read the following paper in which the propensity score matching is criticized.

King, G. and Nielsen, R.(2019)
Why Propensity Scores Should Not Be Used for Matching
Political Analysis, 27(4), 435-454.
https://gking.harvard.edu/publications/why-propensity-scores-should-not-be-used-formatching

Also in the PDF document which is installed with the propensity score mathing add-in, I also introduced Prof. Stephen Senn's critics.