Multicollinearity

[jimjohn]

Hi, just to follow up on this, what if I don't notice any condition indexes that are large (ie none of them are over 20 or 30). Then, in this case, can I say there isn't significant collinearity? Thanks.

SR Millis wrote

VIF (and tolerance) have limitations: the inability to distinguish among several coexisting near-dependencies and the lack of a meaningful guideline to differentiate high VIF from low.

To diagnose collinearity, it is much better to first use the condition indexes: pick out those that are large, say >20 or >30. For those large condition indexes, see if there are large variance-decomposition proportions (> .50) associated with each high condition index: this identifies those variables that have high collinearity.


Scott R Millis, PhD, MEd, ABPP (CN,CL,RP), CStat
Professor & Director of Research
Dept of Physical Medicine & Rehabilitation
Wayne State University School of Medicine
261 Mack Blvd
Detroit, MI 48201
Email:  [e-mail]
Tel: [phone]
Fax: [phone]


--- On Wed, 7/2/08, [e-mail] <[e-mail]> wrote:

> From: [e-mail] <[e-mail]>
> Subject: Re: Multicollinearity
> To: [e-mail]
> Date: Wednesday, July 2, 2008, 10:16 AM
> Thanks so much! I see that SPSS has collinearity
> diagnostics:
> Tolerance and VIF. Can anyone recommend generally what
> values of
> tolerance and VIF should indicate there is a
> multicollinearity
> problem. I am seeing many different responses in different
> lectures/books. some say a tolerance < .1 or a VIF >
> 10 indicate
> collinearity. others say tolerance < .2 and VIF > 4).
> and then ive
> also seen tolerance < .4. Any ideas? thx.
>
>

=====================
To manage your subscription to SPSSX-L, send a message to
[e-mail] (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

[jimjohn]

I have one example where I know there is a multicollinearity effect because my coefficient comes out positive, when it is supposed to have a negative effect on the dependent variable (in a pairwise correlation with the dependent variable, its sign is negative). However, there are no condition indexes greater than 20 or 30, so that test doesn't find any significant collinearity. Also, the values of VIF and tolerance do not show a significant effect either. Can someone plz explain how come these two tests don't show any collinearity when there should be. Or, any suggestions? Thx!



                                Coefficients(a)
                Unstandardized Coefficients Standardized Coefficients Collinearity Statistics
Model B Std. Error Beta t Sig. Tolerance VIF
1 (Constant) .260 .013 20.486 .000
        %VIRM Pacific (Branch) .013 .049 .028 .259 .797 .661 1.513
        LEADS(@3monthBA1monthBA,11) -.455 .084 -.534 -5.383 .000 .809 1.236
        LEADS(@3monthOISCORRA,9) -.147 .037 -.419 -3.947 .000 .708 1.413
a. Dependent Variable: Pacific (Branch)

                        Collinearity Diagnostics(a)
                                        Variance Proportions
Model Dimension Eigenvalue Condition Index (Constant) %VIRM Pacific (Branch) LEADS(@3monthBA1monthBA,11) LEADS(@3monthOISCORRA,9)
1 1 2.747 1.000 .02 .02 .05 .03
        2 .754 1.908 .04 .01 .00 .66
        3 .424 2.546 .05 .02 .92 .09
        4 .075 6.039 .89 .96 .04 .22
a. Dependent Variable: Pacific (Branch)

SR Millis wrote

VIF (and tolerance) have limitations: the inability to distinguish among several coexisting near-dependencies and the lack of a meaningful guideline to differentiate high VIF from low.

To diagnose collinearity, it is much better to first use the condition indexes: pick out those that are large, say >20 or >30. For those large condition indexes, see if there are large variance-decomposition proportions (> .50) associated with each high condition index: this identifies those variables that have high collinearity.


Scott R Millis, PhD, MEd, ABPP (CN,CL,RP), CStat
Professor & Director of Research
Dept of Physical Medicine & Rehabilitation
Wayne State University School of Medicine
261 Mack Blvd
Detroit, MI 48201
Email:  [e-mail]
Tel: [phone]
Fax: [phone]


--- On Wed, 7/2/08, [e-mail] <[e-mail]> wrote:

> From: [e-mail] <[e-mail]>
> Subject: Re: Multicollinearity
> To: [e-mail]
> Date: Wednesday, July 2, 2008, 10:16 AM
> Thanks so much! I see that SPSS has collinearity
> diagnostics:
> Tolerance and VIF. Can anyone recommend generally what
> values of
> tolerance and VIF should indicate there is a
> multicollinearity
> problem. I am seeing many different responses in different
> lectures/books. some say a tolerance < .1 or a VIF >
> 10 indicate
> collinearity. others say tolerance < .2 and VIF > 4).
> and then ive
> also seen tolerance < .4. Any ideas? thx.
>
>

=====================
To manage your subscription to SPSSX-L, send a message to
[e-mail] (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

[SR Millis-3]

I would then consider determining whether the variable is a suppressor variable.

 Scott R Millis, PhD, MEd, ABPP (CN,CL,RP), CStat
Professor & Director of Research
Dept of Physical Medicine & Rehabilitation
Wayne State University School of Medicine
261 Mack Blvd
Detroit, MI 48201
Email: [hidden email]
Tel: [phone]
Fax: [phone]



----- Original Message ----
From: jimjohn <[hidden email]>
To: [hidden email]
Sent: Wednesday, July 16, 2008 3:59:58 PM
Subject: Re: Multicollinearity

I have one example where I know there is a multicollinearity effect because
my coefficient comes out positive, when it is supposed to have a negative
effect on the dependent variable (in a pairwise correlation with the
dependent variable, its sign is negative). However, there are no condition
indexes greater than 20 or 30, so that test doesn't find any significant
collinearity. Also, the values of VIF and tolerance do not show a
significant effect either. Can someone plz explain how come these two tests
don't show any collinearity when there should be. Or, any suggestions? Thx!



                                Coefficients(a)
                Unstandardized Coefficients            Standardized Coefficients                      Collinearity
Statistics
Model          B      Std. Error      Beta    t      Sig.    Tolerance      VIF
1      (Constant)      .260    .013            20.486  .000
        %VIRM Pacific (Branch)  .013    .049    .028    .259    .797    .661    1.513
        LEADS(@3monthBA1monthBA,11)    -.455  .084    -.534  -5.383  .000    .809    1.236
        LEADS(@3monthOISCORRA,9)        -.147  .037    -.419  -3.947  .000    .708    1.413
a. Dependent Variable: Pacific (Branch)

                        Collinearity Diagnostics(a)
                                        Variance Proportions
Model  Dimension      Eigenvalue      Condition Index (Constant)      %VIRM Pacific (Branch)
LEADS(@3monthBA1monthBA,11)    LEADS(@3monthOISCORRA,9)
1      1      2.747  1.000  .02    .02    .05    .03
        2      .754    1.908  .04    .01    .00    .66
        3      .424    2.546  .05    .02    .92    .09
        4      .075    6.039  .89    .96    .04    .22
a. Dependent Variable: Pacific (Branch)



SR Millis wrote:

>
> VIF (and tolerance) have limitations: the inability to distinguish among
> several coexisting near-dependencies and the lack of a meaningful
> guideline to differentiate high VIF from low.
>
> To diagnose collinearity, it is much better to first use the condition
> indexes: pick out those that are large, say >20 or >30. For those large
> condition indexes, see if there are large variance-decomposition
> proportions (> .50) associated with each high condition index: this
> identifies those variables that have high collinearity.
>
>
> Scott R Millis, PhD, MEd, ABPP (CN,CL,RP), CStat
> Professor & Director of Research
> Dept of Physical Medicine & Rehabilitation
> Wayne State University School of Medicine
> 261 Mack Blvd
> Detroit, MI 48201
> Email:  [hidden email]
> Tel: [phone]
> Fax: [phone]
>
>
> --- On Wed, 7/2/08, [hidden email] <[hidden email]> wrote:
>
>> From: [hidden email] <[hidden email]>
>> Subject: Re: Multicollinearity
>> To: [hidden email]
>> Date: Wednesday, July 2, 2008, 10:16 AM
>> Thanks so much! I see that SPSS has collinearity
>> diagnostics:
>> Tolerance and VIF. Can anyone recommend generally what
>> values of
>> tolerance and VIF should indicate there is a
>> multicollinearity
>> problem. I am seeing many different responses in different
>> lectures/books. some say a tolerance < .1 or a VIF >
>> 10 indicate
>> collinearity. others say tolerance < .2 and VIF > 4).
>> and then ive
>> also seen tolerance < .4. Any ideas? thx.
>>
>>
>
> =====================
> 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
>
>

--
View this message in context: http://www.nabble.com/Multicollinearity-tp18197967p18495470.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

[Ornelas, Fermin-2]

I got tempted to reply. I have not idea how you calculated the condition index (intercept included or not), but looking at the variance proportions (.89 and .96) suggest that both variables are highly collinear. If you did not include the intercept it will hide the problem even more...

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of jimjohn
Sent: Wednesday, July 16, 2008 1:00 PM
To: [hidden email]
Subject: Re: Multicollinearity

I have one example where I know there is a multicollinearity effect because
my coefficient comes out positive, when it is supposed to have a negative
effect on the dependent variable (in a pairwise correlation with the
dependent variable, its sign is negative). However, there are no condition
indexes greater than 20 or 30, so that test doesn't find any significant
collinearity. Also, the values of VIF and tolerance do not show a
significant effect either. Can someone plz explain how come these two tests
don't show any collinearity when there should be. Or, any suggestions? Thx!



                                Coefficients(a)
                Unstandardized Coefficients             Standardized Coefficients                       Collinearity
Statistics
Model           B       Std. Error      Beta    t       Sig.    Tolerance       VIF
1       (Constant)      .260    .013            20.486  .000
        %VIRM Pacific (Branch)  .013    .049    .028    .259    .797    .661    1.513
        LEADS(@3monthBA1monthBA,11)     -.455   .084    -.534   -5.383  .000    .809    1.236
        LEADS(@3monthOISCORRA,9)        -.147   .037    -.419   -3.947  .000    .708    1.413
a. Dependent Variable: Pacific (Branch)

                        Collinearity Diagnostics(a)
                                        Variance Proportions
Model   Dimension       Eigenvalue      Condition Index (Constant)      %VIRM Pacific (Branch)
LEADS(@3monthBA1monthBA,11)     LEADS(@3monthOISCORRA,9)
1       1       2.747   1.000   .02     .02     .05     .03
        2       .754    1.908   .04     .01     .00     .66
        3       .424    2.546   .05     .02     .92     .09
        4       .075    6.039   .89     .96     .04     .22
a. Dependent Variable: Pacific (Branch)



SR Millis wrote:

>
> VIF (and tolerance) have limitations: the inability to distinguish among
> several coexisting near-dependencies and the lack of a meaningful
> guideline to differentiate high VIF from low.
>
> To diagnose collinearity, it is much better to first use the condition
> indexes: pick out those that are large, say >20 or >30. For those large
> condition indexes, see if there are large variance-decomposition
> proportions (> .50) associated with each high condition index: this
> identifies those variables that have high collinearity.
>
>
> Scott R Millis, PhD, MEd, ABPP (CN,CL,RP), CStat
> Professor & Director of Research
> Dept of Physical Medicine & Rehabilitation
> Wayne State University School of Medicine
> 261 Mack Blvd
> Detroit, MI 48201
> Email:  [hidden email]
> Tel: [phone]
> Fax: [phone]
>
>
> --- On Wed, 7/2/08, [hidden email] <[hidden email]> wrote:
>
>> From: [hidden email] <[hidden email]>
>> Subject: Re: Multicollinearity
>> To: [hidden email]
>> Date: Wednesday, July 2, 2008, 10:16 AM
>> Thanks so much! I see that SPSS has collinearity
>> diagnostics:
>> Tolerance and VIF. Can anyone recommend generally what
>> values of
>> tolerance and VIF should indicate there is a
>> multicollinearity
>> problem. I am seeing many different responses in different
>> lectures/books. some say a tolerance < .1 or a VIF >
>> 10 indicate
>> collinearity. others say tolerance < .2 and VIF > 4).
>> and then ive
>> also seen tolerance < .4. Any ideas? thx.
>>
>>
>
> =====================
> 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
>
>

--
View this message in context: http://www.nabble.com/Multicollinearity-tp18197967p18495470.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

NOTICE: This e-mail (and any attachments) may contain PRIVILEGED OR CONFIDENTIAL information and is intended only for the use of the specific individual(s) to whom it is addressed.  It may contain information that is privileged and confidential under state and federal law.  This information may be used or disclosed only in accordance with law, and you may be subject to penalties under law for improper use or further disclosure of the information in this e-mail and its attachments. If you have received this e-mail in error, please immediately notify the person named above by reply e-mail, and then delete the original e-mail.  Thank you.

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

[Zdaniuk, Bozena-2]

Hello, everybody. I know a lot about identifying "bad" items in a scale. But are there any established criteria for dropping redundant items? I have a scale with 18 items and the cronbach alpha is .88 and some interitem correlations are .7 or even .9. Can I use interitem correlations to decide which items are redundant and drop them?
Bozena

Bozena Zdaniuk, Ph.D.
University of Pittsburgh
UCSUR, 6th Fl.
121 University Place
Pittsburgh, PA 15260
Ph.: 412-624-5736
Fax: [phone]
Email: [hidden email]

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

[SR Millis-3]

Rasch analysis of a scale allows you to identify redundant items quite directly by examining the person/item map.



Scott R Millis, PhD, ABPP (CN,CL,RP), CStat, CSci
Professor & Director of Research
Dept of Physical Medicine & Rehabilitation
Dept of Emergency Medicine
Wayne State University School of Medicine
261 Mack Blvd
Detroit, MI 48201
Email:  [hidden email]
Tel: [phone]
Fax: [phone]


--- On Wed, 5/27/09, Zdaniuk, Bozena <[hidden email]> wrote:

> From: Zdaniuk, Bozena <[hidden email]>
> Subject: dropping redundant items
> To: [hidden email]
> Date: Wednesday, May 27, 2009, 1:25 PM
> Hello, everybody. I know a lot about
> identifying "bad" items in a scale. But are there any
> established criteria for dropping redundant items? I have a
> scale with 18 items and the cronbach alpha is .88 and some
> interitem correlations are .7 or even .9. Can I use
> interitem correlations to decide which items are redundant
> and drop them?
> Bozena
>
> Bozena Zdaniuk, Ph.D.
> University of Pittsburgh
> UCSUR, 6th Fl.
> 121 University Place
> Pittsburgh, PA 15260
> Ph.: 412-624-5736
> Fax: [phone]
> Email: [hidden email]
>
> =====================
> 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]

You might try looking at the "alpha if item deleted" and drop items one
at a time that 1) lower the alpha then 2) don't lower the alpha much.

Caveat.  Is this an existing scale that has been used in other research?
Then leave the set of items alone.
Is this a new scale you are developing?  If so, do you have several
hundred cases?

Art Kendall
Social Research Consultants

Zdaniuk, Bozena wrote:

> Hello, everybody. I know a lot about identifying "bad" items in a scale. But are there any established criteria for dropping redundant items? I have a scale with 18 items and the cronbach alpha is .88 and some interitem correlations are .7 or even .9. Can I use interitem correlations to decide which items are redundant and drop them?
> Bozena
>
> Bozena Zdaniuk, Ph.D.
> University of Pittsburgh
> UCSUR, 6th Fl.
> 121 University Place
> Pittsburgh, PA 15260
> Ph.: 412-624-5736
> Fax: [phone]
> Email: [hidden email]
>
> =====================
> 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