SPSSX Discussion

Odd discrepancies between SPSS 6 and 15

Classic

List

Threaded

8 messages Options

Judith Saebel

Odd discrepancies between SPSS 6 and 15

Dear all,

I wonder whether someone has an explanation for the following discrepancies between SPSS 6 and 15.
I was checking some old results of exploratory factor analyses (from around 2001). I was able to replicate the results of one of the analyses, but not the other one. However, when I tried to run the same analysis in the old SPSS 6, I got the exactly the same results as in 2001. I'm not talking about minor differences caused by rounding error.

In the 2001 results, the 2 asked-for factors were correlated at .77, and with the exception of one item, all other indicators loaded substantially on one factor and only weakly on the other one.

Using the same data in spss 15, all items load strongly on the same factor but weakly on the other one, and the 2 factors correlate only at .16.

Any explanation please?

Thanks,

Judith Saebel

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

John McConnell

Re: Odd discrepancies between SPSS 6 and 15

Dear Judith,

Are you using SPSS syntax to generate the comparative factor analyses or
are you doing this through the menus/dialog boxes? If the latter I can
see more reason for potential differences as the default options may
have changed over time. This could also affect syntax (i.e. any options
you don't explicitly type can take differing implicit values).

That would be my first thought ... but may not be the answer.

John

John McConnell
Applied Insights

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Judith Saebel
Sent: 28 October 2007 04:51
To: [hidden email]
Subject: Odd discrepancies between SPSS 6 and 15

Dear all,

I wonder whether someone has an explanation for the following
discrepancies between SPSS 6 and 15.
I was checking some old results of exploratory factor analyses (from
around 2001). I was able to replicate the results of one of the
analyses, but not the other one. However, when I tried to run the same
analysis in the old SPSS 6, I got the exactly the same results as in
2001. I'm not talking about minor differences caused by rounding error.

In the 2001 results, the 2 asked-for factors were correlated at .77, and
with the exception of one item, all other indicators loaded
substantially on one factor and only weakly on the other one.

Using the same data in spss 15, all items load strongly on the same
factor but weakly on the other one, and the 2 factors correlate only at
.16.

Any explanation please?

Thanks,

Judith Saebel

=======
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

ViAnn Beadle

Re: Odd discrepancies between SPSS 6 and 15

In reply to this post by Judith Saebel

Have you looked at the descriptive statistics/correlation matrices for
variables in the analysis to make sure that factor is using the same data?

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Judith Saebel
Sent: Saturday, October 27, 2007 10:51 PM
To: [hidden email]
Subject: Odd discrepancies between SPSS 6 and 15

Dear all,

I wonder whether someone has an explanation for the following discrepancies
between SPSS 6 and 15.
I was checking some old results of exploratory factor analyses (from around
2001). I was able to replicate the results of one of the analyses, but not
the other one. However, when I tried to run the same analysis in the old
SPSS 6, I got the exactly the same results as in 2001. I'm not talking
about minor differences caused by rounding error.

In the 2001 results, the 2 asked-for factors were correlated at .77, and
with the exception of one item, all other indicators loaded substantially on
one factor and only weakly on the other one.

Using the same data in spss 15, all items load strongly on the same factor
but weakly on the other one, and the 2 factors correlate only at .16.

Any explanation please?

Thanks,

Judith Saebel

=======
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

Judith Saebel

Re: Odd discrepancies between SPSS 6 and 15

I have. I've just gone through both outputs. Everything is the same until the program gets to the oblimin rotation. SPSS 6 needs 58 iterations to converge, while SPSS 15 takes only 18. Thereafter things change considerably.

________________________________

From: SPSSX(r) Discussion on behalf of ViAnn Beadle
Sent: Mon 29/10/2007 12:29 AM
To: [hidden email]
Subject: Re: Odd discrepancies between SPSS 6 and 15

Have you looked at the descriptive statistics/correlation matrices for
variables in the analysis to make sure that factor is using the same data?

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Judith Saebel
Sent: Saturday, October 27, 2007 10:51 PM
To: [hidden email]
Subject: Odd discrepancies between SPSS 6 and 15

Dear all,

I wonder whether someone has an explanation for the following discrepancies
between SPSS 6 and 15.
I was checking some old results of exploratory factor analyses (from around
2001). I was able to replicate the results of one of the analyses, but not
the other one. However, when I tried to run the same analysis in the old
SPSS 6, I got the exactly the same results as in 2001. I'm not talking
about minor differences caused by rounding error.

In the 2001 results, the 2 asked-for factors were correlated at .77, and
with the exception of one item, all other indicators loaded substantially on
one factor and only weakly on the other one.

Using the same data in spss 15, all items load strongly on the same factor
but weakly on the other one, and the 2 factors correlate only at .16.

Any explanation please?

Thanks,

Judith Saebel

=======
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

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

Hector Maletta

Re: Odd discrepancies between SPSS 6 and 15

I had understood from your previous message that the initial
(orthogonal) factor solution was different, which was improbable. Now I see
the trouble is with oblique rotation. Perhaps the guys at SPSS have adjusted
the iteration algorithm for the Oblimin rotation in recent versions, and
thus this may be the cause of the difference, though it should not cause so
large a difference as you mention.
Was SPSS 6 still programmed in the age-old spaghetti code from the
mainframe era? Some archaeologist, please.

Hector

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Judith Saebel
Sent: 28 October 2007 12:14
To: [hidden email]
Subject: Re: Odd discrepancies between SPSS 6 and 15

I have. I've just gone through both outputs. Everything is the
same until the program gets to the oblimin rotation. SPSS 6 needs 58
iterations to converge, while SPSS 15 takes only 18. Thereafter things
change considerably.

________________________________

From: SPSSX(r) Discussion on behalf of ViAnn Beadle
Sent: Mon 29/10/2007 12:29 AM
To: [hidden email]
Subject: Re: Odd discrepancies between SPSS 6 and 15

Have you looked at the descriptive statistics/correlation matrices
for
variables in the analysis to make sure that factor is using the
same data?

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On
Behalf Of
Judith Saebel
Sent: Saturday, October 27, 2007 10:51 PM
To: [hidden email]
Subject: Odd discrepancies between SPSS 6 and 15

Dear all,

I wonder whether someone has an explanation for the following
discrepancies
between SPSS 6 and 15.
I was checking some old results of exploratory factor analyses
(from around
2001). I was able to replicate the results of one of the analyses,
but not
the other one. However, when I tried to run the same analysis in
the old
SPSS 6, I got the exactly the same results as in 2001. I'm not
talking
about minor differences caused by rounding error.

In the 2001 results, the 2 asked-for factors were correlated at
.77, and
with the exception of one item, all other indicators loaded
substantially on
one factor and only weakly on the other one.

Using the same data in spss 15, all items load strongly on the same
factor
but weakly on the other one, and the 2 factors correlate only at
.16.

Any explanation please?

Thanks,

Judith Saebel

=======
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

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

Judith Saebel

Re: Odd discrepancies between SPSS 6 and 15

Actually, in 2001, I was most likely using SPSS 9, but because I don't access to it anymore, I had to use SPSS 6 which gave me the same loadings etc. as what I had in my notes.

________________________________

From: SPSSX(r) Discussion on behalf of Hector Maletta
Sent: Mon 29/10/2007 2:13 AM
To: [hidden email]
Subject: Re: Odd discrepancies between SPSS 6 and 15

I had understood from your previous message that the initial
(orthogonal) factor solution was different, which was improbable. Now I see
the trouble is with oblique rotation. Perhaps the guys at SPSS have adjusted
the iteration algorithm for the Oblimin rotation in recent versions, and
thus this may be the cause of the difference, though it should not cause so
large a difference as you mention.
Was SPSS 6 still programmed in the age-old spaghetti code from the
mainframe era? Some archaeologist, please.

Hector

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Judith Saebel
Sent: 28 October 2007 12:14
To: [hidden email]
Subject: Re: Odd discrepancies between SPSS 6 and 15

I have. I've just gone through both outputs. Everything is the
same until the program gets to the oblimin rotation. SPSS 6 needs 58
iterations to converge, while SPSS 15 takes only 18. Thereafter things
change considerably.

________________________________

From: SPSSX(r) Discussion on behalf of ViAnn Beadle
Sent: Mon 29/10/2007 12:29 AM
To: [hidden email]
Subject: Re: Odd discrepancies between SPSS 6 and 15

Have you looked at the descriptive statistics/correlation matrices
for
variables in the analysis to make sure that factor is using the
same data?

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On
Behalf Of
Judith Saebel
Sent: Saturday, October 27, 2007 10:51 PM
To: [hidden email]
Subject: Odd discrepancies between SPSS 6 and 15

Dear all,

I wonder whether someone has an explanation for the following
discrepancies
between SPSS 6 and 15.
I was checking some old results of exploratory factor analyses
(from around
2001). I was able to replicate the results of one of the analyses,
but not
the other one. However, when I tried to run the same analysis in
the old
SPSS 6, I got the exactly the same results as in 2001. I'm not
talking
about minor differences caused by rounding error.

In the 2001 results, the 2 asked-for factors were correlated at
.77, and
with the exception of one item, all other indicators loaded
substantially on
one factor and only weakly on the other one.

Using the same data in spss 15, all items load strongly on the same
factor
but weakly on the other one, and the 2 factors correlate only at
.16.

Any explanation please?

Thanks,

Judith Saebel

=======
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

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

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

Lemon, John S.

Re: Odd discrepancies between SPSS 6 and 15

In reply to this post by Hector Maletta

I could really show my age and ask "which version 6 of SPSS?" as there was a version 6 before SPSS-X - I remember using it in the 1970s !!

Any way back to relevant question I am fairly certain that the latest version of 6 was using the old 'mainframe' style of coding as if I remember correctly it was the last version we had which run under Solaris etc. After that it moved to 'server' versions !!

Perhaps we should ask someone at the Directions meeting which starts in Orlando today !!

Best Wishes

John S. Lemon
DIT - Student Liaison Officer
University of Aberdeen
Edward Wright Building: Room G51
Tel: +44 1224 273350
Fax: +44 1224 273372

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Hector Maletta
Sent: 28 October 2007 15:44
To: [hidden email]
Subject: Re: Odd discrepancies between SPSS 6 and 15

I had understood from your previous message that the initial
(orthogonal) factor solution was different, which was improbable. Now I see
the trouble is with oblique rotation. Perhaps the guys at SPSS have adjusted
the iteration algorithm for the Oblimin rotation in recent versions, and
thus this may be the cause of the difference, though it should not cause so
large a difference as you mention.
Was SPSS 6 still programmed in the age-old spaghetti code from the
mainframe era? Some archaeologist, please.

Hector

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

Judith Saebel

Re: Odd discrepancies between SPSS 6 and 15

In reply to this post by Hector Maletta

Dear all,

Fixing delta to .004 did the trick. Comparing the plots of factor rotations from SPSS 6 and 15 gave me the idea. It was then a matter of trial and error until I arrived at an (almost) identical solution to the one from 2001. The differences were minor - a handful of loadings that differed by .01 from the original. And the number of iterations in the oblim rotation went up too - from 18 to 50, which is much closer to the 58 iterations in SPSS 6.

Strangely enough, running another EFA in SPSS 15 produced identical results to those from 2001.

Thanks to all who took time to respond to my question.

Sincerely,

Judith Saebel

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