Pooling PCA results from multiply imputed data

[Joost van Ginkel]

Dear list,

A while ago I posted a message in which I announced a paper discussing the pooling of ANOVA results for multiply imputed data sets. Now recently another paper of mine got published discussing a solution for Principal component analysis. Currently SPSS does not pool the results of PCA for multiply imputed data and averaging the component loadings manually is not a good option, as argued in the paper. The reference for the article is:

Van Ginkel, J.R. & Kroonenberg, P.M. (2014) Using Generalized Procrustes analysis for multiple imputation in principal component analysis. Journal of Classification, 31, 242-269. doi:10.1007/s00357-014-9154-y

I think it would be very helpful for users if this approach got implemented in future versions of SPSS.

Best regards,

Joost van Ginkel

===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD

[Art Kendall]

I'll take a look at that. Sounds interesting.  What kinds of data have you been looking at with your approach?

If you get a chance try running an INDSCAL with each of the separate correlation matrices.
It would be informative to know how the common space compared to the averaged correlations vs the method you propose.

However, as a social scientist I have been more interested in the common variance rather that the  total variance.  (I.e., I usually treat the item specific variance as uninteresting.)

Have you explored your approach with other forms of factor analysis? especially PAF?

[Joost van Ginkel]

Dear Art,

See my answers below (indicated by J:).
________________________________________
From: SPSSX(r) Discussion [[hidden email]] on behalf of Art Kendall [[hidden email]]
Sent: Friday, July 25, 2014 1:49 AM
To: [hidden email]
Subject: Re: Pooling PCA results from multiply imputed data

I'll take a look at that. Sounds interesting.  What kinds of data have you
been looking at with your approach?

J: Could you be more specific on what you mean by "kinds of data"? I looked at simulated data based on existing questionnaire data, although I don't know if that is what you mean.

If you get a chance try running an INDSCAL with each of the separate
correlation matrices.

J: I haven't heard of this procedure. What is it?

It would be informative to know how the common space compared to the
averaged correlations vs the method you propose.

However, as a social scientist I have been more interested in the common
variance rather that the  total variance.  (I.e., I usually treat the item
specific variance as uninteresting.)

Have you explored your approach with other forms of factor analysis?
especially PAF?

J: No I haven't but if I recall correctly I think the paper says that this procedure may be used for PAF as well.

Best regards,

Joost




-----
Art Kendall
Social Research Consultants
--
View this message in context: http://spssx-discussion.1045642.n5.nabble.com/Pooling-PCA-results-from-multiply-imputed-data-tp5726817p5726819.html
Sent from the SPSSX Discussion mailing list archive at Nabble.com.

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD