question about parallel analysis

[Zdaniuk, Bozena-3]

hello everyone,

I have ran the SPSS syntax for parallel analysis developed by Brian O'Connor

(https://people.ok.ubc.ca/brioconn/nfactors/parallel.sps) to check how many factors I should retain. I adjusted the number of cases and variables (n=1848, var#=66) to match my raw data file. The list of eigen values I received in the parallel analysis are very small, starting with .427 so I would need to retain 12 factors if I want to follow the rules of using the parallel analysis to determine factor retention. Since only 4 or five factors in my data are interpretable and make sense from the theoretical point of view, here is my question:

1. Is the parallel analysis useful for larger samples? Does anyone know of anything written about the usefulness of it when determining factor retention on a large data set?

thanks so much,

bozena

---

From: SPSSX(r) Discussion <[hidden email]> on behalf of smalik <[hidden email]>
Sent: Monday, August 3, 2020 7:27 AM
To: [hidden email]
Subject: Re: Help write a syntax - specific question

 

Dear Bruce

Thank you so much for your help. It worked.

AGGREGATE
  /OUTFILE=* MODE=ADDVARIABLES OVERWRITE=YES
  /BREAK=StudentID
  /MinCode=MIN(Code)
  /MaxCode=Max(Code)
  /NumRecs=NU.

IF MinCode EQ MaxCode Type = Code.
IF MinCode NE MaxCode Type = "SOME".
FREQUENCIES Type.
DELETE VARIABLES MinCode MaxCode.



--
Sent from: http://spssx-discussion.1045642.n5.nabble.com/

SPSSX Discussion | Mailing List Archive

spssx-discussion.1045642.n5.nabble.com

SPSSX Discussion forum and mailing list archive.

=====================
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

[Art Kendall]

I suggest (1) that you search the archives of this list for "parallel
analysis" and (2) start new thread since many people will not see this post
which comes up a part of another thread.

There are many considerations in Factor analysis. response scale of items,
previous use of the same items in earlier research, etc.





-----
Art Kendall
Social Research Consultants
--
Sent from: http://spssx-discussion.1045642.n5.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

[Anthony Babinec]

Presumably, you have considered some

of the “factorability” statistics(e.g., MSA,

initial communalities) for the information

they provide.

 

If you are interested in factor analysis and

not PCA, be sure to specify kind=2

 

* Specify the desired kind of parallel analysis, where:

  1 = principal components analysis

  2 = principal axis/common factor analysis.

compute kind = 2 .

 

The eigenanalysis is based on eigenvalues of the

reduced correlation matrix. Eigenvalues can be

positive or negative. You might use an eigenvalue

greater than 0 rule, but that can lead to too many

factors retained. If based on raw data, you can do the

bootstrap analysis supported in the routine. Note

the magnitudes of the eigenvalues relative to the

corresponding 90^th percentile values based on random

data. You can also produce “by hand” a scree plot

and look for an “elbow.” These and other ideas are

discussed in “Exploratory Factor Analysis” by

Leandre Fabrigar and Duane Wegener.

 

Tony Babinec

===================== 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