Please send me the files. I'll look into it after I finish my current extension project, conditional inference trees, which looks to be pretty exciting.
As for how easy to add MATRIX OUT to MVA, while I don't know that code, I can think of reasons why it might be more work than you think. Nevertheless, it's always an issue of priorities for scarce resources.
--
Original Message:
Sent: 10/25/2024 10:00:00 AM
From: Bruce Weaver
Subject: RE: Calculating Mean Values with Pairwise Deletion of Missing Values for a Set of Variables
Hi Jon. As you know from previous such conversations, I always wish that feature like this were incorporated in official commands. And in this particular case, it seems to me that it would not be that difficult to add /MATRIX OUT to MVA, given that it has already been done for CORRELATIONS. (But there may be complications I am not aware of.) In the meanwhile, I suspect many users would find an extension command easier to use than my macros, especially if it had a GUI. So if you're interested in tackling it, go ahead I say!
PS- The Google Sites links at the end of that article no longer work. But I can send you the files via email if you ever want them.
------------------------------
Bruce Weaver
------------------------------
Original Message:
Sent: Fri October 25, 2024 09:41 AM
From: Jon Peck
Subject: Calculating Mean Values with Pairwise Deletion of Missing Values for a Set of Variables
Hi Bruce,
Not being one of the "powers that be", (maybe powers that were), I can't answer your MATRIX OUT question, but it seems from your article that generating the appropriate matrix is actually pretty easy using your macros. If that's still too hard, how about an extension command to do this?
--
Original Message:
Sent: 10/25/2024 7:11:00 AM
From: Bruce Weaver
Subject: RE: Calculating Mean Values with Pairwise Deletion of Missing Values for a Set of Variables
Hello Dariush. You have not said why you want to use pairwise deletion, and I wonder if you are aware of the shortcomings of that approach. For a brief overview, search for <pairwise> and <factor analysis> in this nice Annual Review chapter by John Graham:
- https://imaging.mrc-cbu.cam.ac.uk/statswiki/FAQ/emalgm?action=AttachFile&do=get&target=graham.pdf
See also my 2014 article with Hillary Maxwell for info on how to use a matrix of EM covariances (or correlations) as input for factor analysis (the method Graham recommends over pairwise deletion):
- https://www.tqmp.org/RegularArticles/vol10-2/p143/p143.pdf
Finally, a question for the powers that be at IBM-SPSS: Why does MVA still not have a /MATRIX OUT sub-command? As much as I appreciate any citations of that 2014 article, I wish it was much easier for users to use EM correlation (or covariance) matrices as input to FACTOR.
Cheers,
Bruce
------------------------------
Bruce Weaver
Original Message:
Sent: Thu October 17, 2024 04:00 PM
From: Dariush Azimi
Subject: Calculating Mean Values with Pairwise Deletion of Missing Values for a Set of Variables
Hi there
I have conducted a Factor Analysis using pairwise deletion of missing values. The resulting factors consist of several variables/items. I need to calculate the mean values as well as standard deviation with pairwise deletion of missing values of the items of each factor, which are all likert-scaled (4). I did achieve doing this with listwise deletion of missing values, but cant find a way to do it with pairwise deletion. The SPSS Version Im using is IBM SPSS Statistics 29.0
Cheers everyone and thank you in advance.
It is probably a simple question for you guys. You would help me loads if someone could tell me how it works.
------------------------------
Dariush Azimi
------------------------------