SPSSX Discussion

Path analysis and percentage of variance explained

Classic

List

Threaded

7 messages Options

Evelyn Yu

Path analysis and percentage of variance explained

Dear all,

I wonder if any of the users from this list know whether it is possible to
quantify the percentage of variance explained by each predictor variable
onto the dependent variable from a Path analysis/just a simple multiple
linear regression using AMOS/SPSS.

In my understanding, the absolute value of the standardized estimates would
allow me to compare the relative magnitude that each predictor has onto the
dependent variable, but not the exact value. May I obtain such an estimate
about each predictor variable by looking at the R-square change after each
predictor variable has been entered into the multiple linear regression
model in SPSS?

Many thanks for your kind help and advice in advance.

Sincerely,

Evelyn Yu
[hidden email]
Department of Medicine & Therapeutics
The Chinese University of Hong Kong

David Greenberg

Re: Path analysis and percentage of variance explained

In regression, unless all the independent variables are mutually uncorrelated, it is not possible to partition the explained variance in a manner that assigns to each independent variable its unique contribution to the total amount of variance explained. David Greenberg, Sociology Department, New York University

---- Original Message -----
From: Evelyn Yu <[hidden email]>
Date: Monday, June 25, 2007 5:59 am
Subject: Path analysis and percentage of variance explained
To: [hidden email]

> Dear all,
>
> I wonder if any of the users from this list know whether it is
> possible to
> quantify the percentage of variance explained by each predictor variable
> onto the dependent variable from a Path analysis/just a simple multiple
> linear regression using AMOS/SPSS.
>
> In my understanding, the absolute value of the standardized estimates
> would
> allow me to compare the relative magnitude that each predictor has
> onto the
> dependent variable, but not the exact value. May I obtain such an estimate
> about each predictor variable by looking at the R-square change after
> each
> predictor variable has been entered into the multiple linear regression
> model in SPSS?
>
> Many thanks for your kind help and advice in advance.
>
> Sincerely,
>
> Evelyn Yu
> [hidden email]
> Department of Medicine & Therapeutics
> The Chinese University of Hong Kong

Ornelas, Fermin

FW: Re: Path analysis and percentage of variance explained

In reply to this post by Evelyn Yu

-----Original Message-----
From: Ornelas, Fermin
Sent: Monday, June 25, 2007 2:55 PM
To: 'David Greenberg'
Subject: RE: Re: Path analysis and percentage of variance explained

But if you want to judge the variable importance in its power to
contribute to the dependent variable you could use the standard error or
t-test for each predictor to judge how important a variable according to
those values, the higher the t-test the more important the variable is
in its contribution on the behavior of the response variable.

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
David Greenberg
Sent: Monday, June 25, 2007 1:27 PM
To: [hidden email]
Subject: Re: Path analysis and percentage of variance explained

In regression, unless all the independent variables are mutually
uncorrelated, it is not possible to partition the explained variance in
a manner that assigns to each independent variable its unique
contribution to the total amount of variance explained. David Greenberg,
Sociology Department, New York University

---- Original Message -----
From: Evelyn Yu <[hidden email]>
Date: Monday, June 25, 2007 5:59 am
Subject: Path analysis and percentage of variance explained
To: [hidden email]

> Dear all,
>
> I wonder if any of the users from this list know whether it is
> possible to
> quantify the percentage of variance explained by each predictor
variable
> onto the dependent variable from a Path analysis/just a simple
multiple
> linear regression using AMOS/SPSS.
>
> In my understanding, the absolute value of the standardized estimates
> would
> allow me to compare the relative magnitude that each predictor has
> onto the
> dependent variable, but not the exact value. May I obtain such an
estimate
> about each predictor variable by looking at the R-square change after
> each
> predictor variable has been entered into the multiple linear
regression

> model in SPSS?
>
> Many thanks for your kind help and advice in advance.
>
> Sincerely,
>
> Evelyn Yu
> [hidden email]
> Department of Medicine & Therapeutics
> The Chinese University of Hong Kong

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.

Swank, Paul R

Re: Path analysis and percentage of variance explained

I would agree with David on this. Measuring the im,portance of one
variable among many to the dependent variable is a difficult
proposition. Many people think that the t or p values tell you that but
they don't. They might in a perfect world where all the variables are
measured to the same degree of precision, there were no outliers, no
collinearity, and all pertinent variables were included in the analysis.
But of course, this is never so.

Paul R. Swank, Ph.D. Professor
Director of Reseach
Children's Learning Institute
University of Texas Health Science Center-Houston

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Ornelas, Fermin
Sent: Monday, June 25, 2007 4:56 PM
To: [hidden email]
Subject: FW: Re: Path analysis and percentage of variance explained

-----Original Message-----
From: Ornelas, Fermin
Sent: Monday, June 25, 2007 2:55 PM
To: 'David Greenberg'
Subject: RE: Re: Path analysis and percentage of variance explained

But if you want to judge the variable importance in its power to
contribute to the dependent variable you could use the standard error or
t-test for each predictor to judge how important a variable according to
those values, the higher the t-test the more important the variable is
in its contribution on the behavior of the response variable.

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
David Greenberg
Sent: Monday, June 25, 2007 1:27 PM
To: [hidden email]
Subject: Re: Path analysis and percentage of variance explained

In regression, unless all the independent variables are mutually
uncorrelated, it is not possible to partition the explained variance in
a manner that assigns to each independent variable its unique
contribution to the total amount of variance explained. David Greenberg,
Sociology Department, New York University

---- Original Message -----
From: Evelyn Yu <[hidden email]>
Date: Monday, June 25, 2007 5:59 am
Subject: Path analysis and percentage of variance explained
To: [hidden email]

> Dear all,
>
> I wonder if any of the users from this list know whether it is
> possible to quantify the percentage of variance explained by each
> predictor
variable
> onto the dependent variable from a Path analysis/just a simple
multiple
> linear regression using AMOS/SPSS.
>
> In my understanding, the absolute value of the standardized estimates
> would allow me to compare the relative magnitude that each predictor
> has onto the dependent variable, but not the exact value. May I
> obtain such an
estimate
> about each predictor variable by looking at the R-square change after
> each predictor variable has been entered into the multiple linear
regression

> model in SPSS?
>
> Many thanks for your kind help and advice in advance.
>
> Sincerely,
>
> Evelyn Yu
> [hidden email]
> Department of Medicine & Therapeutics
> The Chinese University of Hong Kong

Ornelas, Fermin

Re: Path analysis and percentage of variance explained

Presumably one must first do an exhaustive exploratory analysis so as to
get the results as robust as possible. But at the end of the day it ends
up being a somewhat subjective call.

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Swank, Paul R
Sent: Monday, June 25, 2007 3:11 PM
To: [hidden email]
Subject: Re: Path analysis and percentage of variance explained

I would agree with David on this. Measuring the im,portance of one
variable among many to the dependent variable is a difficult
proposition. Many people think that the t or p values tell you that but
they don't. They might in a perfect world where all the variables are
measured to the same degree of precision, there were no outliers, no
collinearity, and all pertinent variables were included in the analysis.
But of course, this is never so.

Paul R. Swank, Ph.D. Professor
Director of Reseach
Children's Learning Institute
University of Texas Health Science Center-Houston

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Ornelas, Fermin
Sent: Monday, June 25, 2007 4:56 PM
To: [hidden email]
Subject: FW: Re: Path analysis and percentage of variance explained

-----Original Message-----
From: Ornelas, Fermin
Sent: Monday, June 25, 2007 2:55 PM
To: 'David Greenberg'
Subject: RE: Re: Path analysis and percentage of variance explained

But if you want to judge the variable importance in its power to
contribute to the dependent variable you could use the standard error or
t-test for each predictor to judge how important a variable according to
those values, the higher the t-test the more important the variable is
in its contribution on the behavior of the response variable.

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
David Greenberg
Sent: Monday, June 25, 2007 1:27 PM
To: [hidden email]
Subject: Re: Path analysis and percentage of variance explained

In regression, unless all the independent variables are mutually
uncorrelated, it is not possible to partition the explained variance in
a manner that assigns to each independent variable its unique
contribution to the total amount of variance explained. David Greenberg,
Sociology Department, New York University

---- Original Message -----
From: Evelyn Yu <[hidden email]>
Date: Monday, June 25, 2007 5:59 am
Subject: Path analysis and percentage of variance explained
To: [hidden email]

> Dear all,
>
> I wonder if any of the users from this list know whether it is
> possible to quantify the percentage of variance explained by each
> predictor
variable
> onto the dependent variable from a Path analysis/just a simple
multiple
> linear regression using AMOS/SPSS.
>
> In my understanding, the absolute value of the standardized estimates
> would allow me to compare the relative magnitude that each predictor
> has onto the dependent variable, but not the exact value. May I
> obtain such an
estimate
> about each predictor variable by looking at the R-square change after
> each predictor variable has been entered into the multiple linear
regression

> model in SPSS?
>
> Many thanks for your kind help and advice in advance.
>
> Sincerely,
>
> Evelyn Yu
> [hidden email]
> Department of Medicine & Therapeutics
> The Chinese University of Hong Kong

David Greenberg

Re: FW: Re: Path analysis and percentage of variance explained

In reply to this post by Ornelas, Fermin

You can certainly measure the strength of the coefficients, though significance tests are not measures of strength. However, the questioner asked about assigning to each variable the fraction of the explained variance due to each. That cannot be done uniquely as long as the predictors are correlated. Of course, if one specifies the order in which variables are entered, one can see how much additional variance each one explains. David Greenberg, Sociology Department, New York University

----- Original Message -----
From: "Ornelas, Fermin" <[hidden email]>
Date: Monday, June 25, 2007 5:56 pm
Subject: FW: Re: Path analysis and percentage of variance explained
To: [hidden email]

> -----Original Message-----
> From: Ornelas, Fermin
> Sent: Monday, June 25, 2007 2:55 PM
> To: 'David Greenberg'
> Subject: RE: Re: Path analysis and percentage of variance explained
>
> But if you want to judge the variable importance in its power to
> contribute to the dependent variable you could use the standard error
> or
> t-test for each predictor to judge how important a variable according
> to
> those values, the higher the t-test the more important the variable is
> in its contribution on the behavior of the response variable.
>
>
> -----Original Message-----
> From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf
> Of
> David Greenberg
> Sent: Monday, June 25, 2007 1:27 PM
> To: [hidden email]
> Subject: Re: Path analysis and percentage of variance explained
>
> In regression, unless all the independent variables are mutually
> uncorrelated, it is not possible to partition the explained variance in
> a manner that assigns to each independent variable its unique
> contribution to the total amount of variance explained. David Greenberg,
> Sociology Department, New York University
>
> ---- Original Message -----
> From: Evelyn Yu <[hidden email]>
> Date: Monday, June 25, 2007 5:59 am
> Subject: Path analysis and percentage of variance explained
> To: [hidden email]
>
>
> > Dear all,
> >
> > I wonder if any of the users from this list know whether it is
> > possible to
> > quantify the percentage of variance explained by each predictor
> variable
> > onto the dependent variable from a Path analysis/just a simple
> multiple
> > linear regression using AMOS/SPSS.
> >
> > In my understanding, the absolute value of the standardized estimates
> > would
> > allow me to compare the relative magnitude that each predictor has
> > onto the
> > dependent variable, but not the exact value. May I obtain such an
> estimate
> > about each predictor variable by looking at the R-square change after
> > each
> > predictor variable has been entered into the multiple linear
> regression
> > model in SPSS?
> >
> > Many thanks for your kind help and advice in advance.
> >
> > Sincerely,
> >
> > Evelyn Yu
> > [hidden email]
> > Department of Medicine & Therapeutics
> > The Chinese University of Hong Kong
>
> 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.

Evelyn Yu

Re: Path analysis and percentage of variance explained

In reply to this post by Ornelas, Fermin

Dear all,

Thank you for your comments, as well as Dr. Greenberg's reminder on the
importance of correlation in regression analysis. They are the most
helpful.

Evelyn

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Ornelas, Fermin
Sent: Tuesday, June 26, 2007 6:30 AM
To: [hidden email]
Subject: Re: Path analysis and percentage of variance explained

Presumably one must first do an exhaustive exploratory analysis so as to
get the results as robust as possible. But at the end of the day it ends
up being a somewhat subjective call.

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Swank, Paul R
Sent: Monday, June 25, 2007 3:11 PM
To: [hidden email]
Subject: Re: Path analysis and percentage of variance explained

I would agree with David on this. Measuring the im,portance of one
variable among many to the dependent variable is a difficult
proposition. Many people think that the t or p values tell you that but
they don't. They might in a perfect world where all the variables are
measured to the same degree of precision, there were no outliers, no
collinearity, and all pertinent variables were included in the analysis.
But of course, this is never so.

Paul R. Swank, Ph.D. Professor
Director of Reseach
Children's Learning Institute
University of Texas Health Science Center-Houston

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Ornelas, Fermin
Sent: Monday, June 25, 2007 4:56 PM
To: [hidden email]
Subject: FW: Re: Path analysis and percentage of variance explained

-----Original Message-----
From: Ornelas, Fermin
Sent: Monday, June 25, 2007 2:55 PM
To: 'David Greenberg'
Subject: RE: Re: Path analysis and percentage of variance explained

But if you want to judge the variable importance in its power to
contribute to the dependent variable you could use the standard error or
t-test for each predictor to judge how important a variable according to
those values, the higher the t-test the more important the variable is
in its contribution on the behavior of the response variable.

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
David Greenberg
Sent: Monday, June 25, 2007 1:27 PM
To: [hidden email]
Subject: Re: Path analysis and percentage of variance explained

In regression, unless all the independent variables are mutually
uncorrelated, it is not possible to partition the explained variance in
a manner that assigns to each independent variable its unique
contribution to the total amount of variance explained. David Greenberg,
Sociology Department, New York University

---- Original Message -----
From: Evelyn Yu <[hidden email]>
Date: Monday, June 25, 2007 5:59 am
Subject: Path analysis and percentage of variance explained
To: [hidden email]

> Dear all,
>
> I wonder if any of the users from this list know whether it is
> possible to quantify the percentage of variance explained by each
> predictor
variable
> onto the dependent variable from a Path analysis/just a simple
multiple
> linear regression using AMOS/SPSS.
>
> In my understanding, the absolute value of the standardized estimates
> would allow me to compare the relative magnitude that each predictor
> has onto the dependent variable, but not the exact value. May I
> obtain such an
estimate
> about each predictor variable by looking at the R-square change after
> each predictor variable has been entered into the multiple linear
regression

> model in SPSS?
>
> Many thanks for your kind help and advice in advance.
>
> Sincerely,
>
> Evelyn Yu
> [hidden email]
> Department of Medicine & Therapeutics
> The Chinese University of Hong Kong