Problems with regression analyses

[J]

Hello altogether,

I have to conduct regression analyses (I have 4 predictors, whereby 2 are investigated with one regression analysis in each case).  They showed, that the regression analysis were significant for an interaction in one case and for a main effect in another case. This makes sense to my theory. However the intercorrelation of the variables seems strange to me. The IV do correlate with each other (.16, .21* and .29**).  None of the IV however correlates with the DV. How can this be explained? I hope you can help me, what to do and what is wrong.

Best!  

[Maguin, Eugene]

J,

Since you are on nabble, please post the regression output (syntax plus coefficients table) for both regressions and the supporting correlation matrix for all variables involved in the two regressions. It's hard to know what really going on without seeing the data results.

Gene Maguin

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of J
Sent: Thursday, August 02, 2012 2:08 AM
To: [hidden email]
Subject: Problems with regression analyses

Hello altogether,

I have to conduct regression analyses (I have 4 predictors, whereby 2 are investigated with one regression analysis in each case).  They showed, that the regression analysis were significant for an interaction in one case and for a main effect in another case. This makes sense to my theory. However the intercorrelation of the variables seems strange to me. The IV do correlate with each other (.16, .21* and .29**).  None of the IV however correlates with the DV. How can this be explained? I hope you can help me, what to do and what is wrong.

Best!



--
View this message in context: http://spssx-discussion.1045642.n5.nabble.com/Problems-with-regression-analyses-tp5714564.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

[J]

Hello Eugene,

I can do this on monday.

Best,
J

[J]

Eugene, I attached the file with the coefficients of my three regressions and the correlation matrix. I hope you can see it. Thanks. JNabble.spv

[Maguin, Eugene]

J

That file, nabble.spv, is zip file with a combination of .bin and xml files. I had expected something that opened in spss; neither of those file types do. Gene Maguin

 

From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of J
Sent: Monday, August 06, 2012 2:28 AM
To: [hidden email]
Subject: Re: Problems with regression analyses

 

Eugene, I attached the file with the coefficients of my three regressions and the correlation matrix. I hope you can see it. Thanks. JNabble.spv

---

View this message in context: Re: Problems with regression analyses
Sent from the SPSSX Discussion mailing list archive at Nabble.com.

[David Marso]

What you attached is not readable by any software on my system.
Best to post a MATRIX file and the syntax used in the regressions.
CORRELATION VARIABLES= <variable list of all variables in all regressions> / MATRIX= OUT (filename).

J wrote

Eugene,

I attached the file with the coefficients of my three regressions and the correlation matrix. I hope you can see it.

Thanks.
JNabble.spv

[Bruce Weaver]

When I clicked the link, an output file (.spv extension) opened with no problems.  Here it is exported to PDF.  

Nabble.pdf

Maguin, Eugene wrote

J
That file, nabble.spv, is zip file with a combination of .bin and xml files. I had expected something that opened in spss; neither of those file types do. Gene Maguin

From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of J
Sent: Monday, August 06, 2012 2:28 AM
To: [hidden email]
Subject: Re: Problems with regression analyses

Eugene, I attached the file with the coefficients of my three regressions and the correlation matrix. I hope you can see it. Thanks. JNabble.spv<http://spssx-discussion.1045642.n5.nabble.com/file/n5714592/Nabble.spv>
________________________________
View this message in context: Re: Problems with regression analyses<http://spssx-discussion.1045642.n5.nabble.com/Problems-with-regression-analyses-tp5714564p5714592.html>
Sent from the SPSSX Discussion mailing list archive<http://spssx-discussion.1045642.n5.nabble.com/> at Nabble.com.

[Rich Ulrich]

I have looked at the .pdf version of the output that Bruce provided.
Google let me know that "Standardfehler" is a word meaning
standard error, so that the trailing r in the next line was not an "r"
of some kind.

Your worry, in this first message, is that "None of the IV ... correlates
with the DV."   But that is clearly wrong.  The regression test with
one variable is, exactly, a test on the correlation coefficient --
which is about .25  for both of the regressions, judging from
the standardized betas.

Perhaps you were looking at wrong place to find correlations.

--
Rich Ulrich

> Date: Wed, 1 Aug 2012 23:08:13 -0700

> From: [hidden email]
> Subject: Problems with regression analyses
> To: [hidden email]
>
> Hello altogether,
>
> I have to conduct regression analyses (I have 4 predictors, whereby 2 are
> investigated with one regression analysis in each case). They showed, that
> the regression analysis were significant for an interaction in one case and
> for a main effect in another case. This makes sense to my theory. However
> the intercorrelation of the variables seems strange to me. The IV do
> correlate with each other (.16, .21* and .29**). None of the IV however
> correlates with the DV. How can this be explained? I hope you can help me,
> what to do and what is wrong.
>
> Best!
> ...

[David Marso]

It is pretty useless from a diagnostic point of view.  For example where are the requested intercorrelations between the predictors?  I believe the OP needs to be both more concise WRT provided output and more specific WRT the supposed issues.  Also difficult to mentally parse the german output.

OP: Please remove ALL variables not used in the regressions from any correlation output and provide the SYNTAX.  Preferably attach the information (MATRIX system file) I requested earlier rather than output.
We require the complete correlation MATRIX (with Means and SDs) of *ONLY* the variables involved, NOT any extraneous ones.  I typically dedicate a maximum of about 10 minutes on any question in this forum and it would take more than that simply to isolate the specific issues from this output.  
--

Bruce Weaver wrote

When I clicked the link, an output file (.spv extension) opened with no problems.  Here it is exported to PDF.  

Nabble.pdf

Maguin, Eugene wrote

J
That file, nabble.spv, is zip file with a combination of .bin and xml files. I had expected something that opened in spss; neither of those file types do. Gene Maguin

From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of J
Sent: Monday, August 06, 2012 2:28 AM
To: [hidden email]
Subject: Re: Problems with regression analyses

Eugene, I attached the file with the coefficients of my three regressions and the correlation matrix. I hope you can see it. Thanks. JNabble.spv<http://spssx-discussion.1045642.n5.nabble.com/file/n5714592/Nabble.spv>
________________________________
View this message in context: Re: Problems with regression analyses<http://spssx-discussion.1045642.n5.nabble.com/Problems-with-regression-analyses-tp5714564p5714592.html>
Sent from the SPSSX Discussion mailing list archive<http://spssx-discussion.1045642.n5.nabble.com/> at Nabble.com.