SPSSX Discussion

linear regression coefficient of zero

Classic

List

Threaded

8 messages Options

Phil Schubert

linear regression coefficient of zero

i'm an inexperienced user trying to complete a dissertation. in a recent linear regression model, the SPSS output indicated that one of my independent variables had a coefficient value of zero, despite showing to be statistically significant. The independent variable was pulled in from from excel and was originally in dollar form with the $ symbol and decimals. From the data editor, i changed the variable to a numeric, scale variable. I also attempted a second regression model, using only this one independent variable that i had trouble with. Again it showed to be significant, but the coefficient value was zero. I assume this has something to do with the way i have the data entered/coded. Any ideas on how to fix this would be appreciated.
--
Phil Schubert
Abilene Christian University
Executive Vice President
ACU Box 29120
Abilene, Texas 79699-9120
325/674-2795

Swank, Paul R

Re: linear regression coefficient of zero

Have you tried rescaling it, say by diving the variable by 100?

Dr. Paul R. Swank,

Professor and Director of Research

Children's Learning Institute

University of Texas Health Science Center-Houston

From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Phil Schubert
Sent: Thursday, March 05, 2009 11:08 AM
To: [hidden email]
Subject: linear regression coefficient of zero

Garry Gelade

Re: linear regression coefficient of zero

When you say “the coefficient”, do you mean the standardized coefficient or the unstandardized coefficient? If it is the unstandardized coefficient that is zero, Paul’s method (re-scaling) will help you. In fact the coefficient is probably not zero, it is .00000something, which you can examine by increasing the number of decimal digits displayed.

If it is the standardized coefficient that is zero, you have a real problem!

Garry Gelade

Business Analytic.

From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Swank, Paul R
Sent: 05 March 2009 17:25
To: [hidden email]
Subject: Re: linear regression coefficient of zero

Have you tried rescaling it, say by diving the variable by 100?

Dr. Paul R. Swank,

Professor and Director of Research

Children's Learning Institute

University of Texas Health Science Center-Houston

__________ NOD32 3911 (20090305) Information __________

This message was checked by NOD32 antivirus system.
http://www.eset.com

John Amora

Re: linear regression coefficient of zero

In reply to this post by Phil Schubert

Hi Phil, Paul and Garry,

Garry is right that it is not zero, its something .00000something. Perhaps your independent is expressed in thousands or millions of dollars. If this is the case, then Paul's suggestions is the solution. Re-scale your independent, say 100 millions instead of 100,000,000 by dividing 1,000, 000; or 100 thousands instead of 100, 000 by dividing 1000; etc.

Cheers,

Johnny

--- On Fri, 3/6/09, Garry Gelade <[hidden email]> wrote:

From: Garry Gelade <[hidden email]..uk>
Subject: Re: linear regression coefficient of zero
To: [hidden email]
Date: Friday, 6 March, 2009, 11:08 AM

When you say “the coefficient”, do you mean the standardized coefficient or the unstandardized coefficient? If it is the unstandardized coefficient that is zero, Paul’s method (re-scaling) will help you. In fact the coefficient is probably not zero, it is .00000something, which you can examine by increasing the number of decimal digits displayed.

If it is the standardized coefficient that is zero, you have a real problem!

Garry Gelade
Business Analytic.

From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Swank, Paul R
Sent: 05 March 2009 17:25
To: [hidden email]
Subject: Re: linear regression coefficient of zero

Have you tried rescaling it, say by diving the variable by 100?

Dr. Paul R. Swank,
Professor and Director of Research
Children's Learning Institute
University of Texas Health Science Center-Houston

From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Phil Schubert
Sent: Thursday, March 05, 2009 11:08 AM
To: [hidden email]
Subject: linear regression coefficient of zero

i'm an inexperienced user trying to complete a dissertation. in a recent linear regression model, the SPSS output indicated that one of my independent variables had a coefficient value of zero, despite showing to be statistically significant. The independent variable was pulled in from from excel and was originally in dollar form with the $ symbol and decimals. From the data editor, i changed the variable to a numeric, scale variable. I also attempted a second regression model, using only this one independent variable that i had trouble with. Again it showed to be significant, but the coefficient value was zero. I assume this has something to do with the way i have the data entered/coded. Any ideas on how to fix this would be appreciated.
--
Phil Schubert
Abilene Christian University
Executive Vice President
ACU Box 29120
Abilene , Texas 79699-9120
325/674-2795

__________ NOD32 3911 (20090305) Information __________

This message was checked by NOD32 antivirus system.
http://www.eset.com

Importing contacts has never been easier.
Bring your friends over to Yahoo! Mail today!

Johnny Amora

Re: linear regression coefficient of zero

In reply to this post by Phil Schubert

Hi Phil, Paul and Garry,

Garry is right that it is not zero, its something .00000something. I suspect that your independent is expressed in thousands or millions of dollars. If this is the case, then Paul's suggestions is the solution. Re-scale your independent, say 100 millions instead of 100,000,000 by dividing 1,000, 000; or 100 thousands instead of 100, 000 by dividing 1000; etc.

Cheers,

Johnny

--- On Fri, 3/6/09, Garry Gelade <[hidden email]> wrote:

From: Garry Gelade <[hidden email]..uk>
Subject: Re: linear regression coefficient of zero
To: [hidden email]
Date: Friday, 6 March, 2009, 11:08 AM

When you say “the coefficient”, do you mean the standardized coefficient or the unstandardized coefficient? If it is the unstandardized coefficient that is zero, Paul’s method (re-scaling) will help you. In fact the coefficient is probably not zero, it is .00000something, which you can examine by increasing the number of decimal digits displayed.

If it is the standardized coefficient that is zero, you have a real problem!

Garry Gelade
Business Analytic.

From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Swank, Paul R
Sent: 05 March 2009 17:25
To: [hidden email]
Subject: Re: linear regression coefficient of zero

Have you tried rescaling it, say by diving the variable by 100?

Dr. Paul R. Swank,
Professor and Director of Research
Children's Learning Institute
University of Texas Health Science Center-Houston

From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Phil Schubert
Sent: Thursday, March 05, 2009 11:08 AM
To: [hidden email]
Subject: linear regression coefficient of zero

i'm an inexperienced user trying to complete a dissertation. in a recent linear regression model, the SPSS output indicated that one of my independent variables had a coefficient value of zero, despite showing to be statistically significant. The independent variable was pulled in from from excel and was originally in dollar form with the $ symbol and decimals. From the data editor, i changed the variable to a numeric, scale variable. I also attempted a second regression model, using only this one independent variable that i had trouble with. Again it showed to be significant, but the coefficient value was zero. I assume this has something to do with the way i have the data entered/coded. Any ideas on how to fix this would be appreciated.
--
Phil Schubert
Abilene Christian University
Executive Vice President
ACU Box 29120
Abilene , Texas 79699-9120
325/674-2795

__________ NOD32 3911 (20090305) Information __________

This message was checked by NOD32 antivirus system.
http://www.eset.com

Try the new Yahoo! Messenger 9.0 It is finally here!

Hector Maletta

Re: linear regression coefficient of zero

In reply to this post by John Amora

The general idea that the scale (the unit of measurement) is the culprit is correct. But the idea of diminishing the size of the unit of measurement of the INDEPENDENT variable is wrong. The effect of an increase of one million dollars is surely higher than the effect of an increase of just one dollar. Thus, expressing incomes in dollars instead of thousands of dollars would REDUCE the absolute value of the coefficient. What is needed is the reverse: INCREASE the unit of the independent (or reduce the unit of the dependent), to get coefficients with a higher absolute value.

Besides, as remarked in another message to this thread, the STANDARDIZED COEFFICIENTS (whose unit is the standard deviation of each variable) should not suffer this problem.

Hector

From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Johnny Amora
Sent: 06 March 2009 08:13
To: [hidden email]
Subject: Re: linear regression coefficient of zero

Hi Phil, Paul and Garry,

Cheers,

Johnny

--- On Fri, 3/6/09, Garry Gelade <[hidden email]> wrote:

From: Garry Gelade <[hidden email]..uk>
Subject: Re: linear regression coefficient of zero
To: [hidden email]
Date: Friday, 6 March, 2009, 11:08 AM

When you say “the coefficient”, do you mean the standardized coefficient or the unstandardized coefficient? If it is the unstandardized coefficient that is zero, Paul’s method (re-scaling) will help you. In fact the coefficient is probably not zero, it is .00000something, which you can examine by increasing the number of decimal digits displayed.

If it is the standardized coefficient that is zero, you have a real problem!

Garry Gelade

Business Analytic.

From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Swank, Paul R
Sent: 05 March 2009 17:25
To: [hidden email]
Subject: Re: linear regression coefficient of zero

Have you tried rescaling it, say by diving the variable by 100?

Dr. Paul R. Swank,

Professor and Director of Research

Children's Learning Institute

University of Texas Health Science Center-Houston

__________ NOD32 3911 (20090305) Information __________

This message was checked by NOD32 antivirus system.
http://www.eset.com

Importing contacts has never been easier.
Bring your friends over to Yahoo! Mail today!

Johnny Amora

Re: linear regression coefficient of zero

In reply to this post by Phil Schubert

The suggestion was just the same as using 1 kilogram, instead of 1000grams. I dont think such transformation in the indepedent is wrong. I dont think this is wrong: 1,000,000 USD= 1Million USD

--- On Sat, 3/7/09, Hector Maletta <[hidden email]> wrote:

From: Hector Maletta <[hidden email]>
Subject: Re: linear regression coefficient of zero
To: [hidden email]
Date: Saturday, 7 March, 2009, 12:06 AM

The general idea that the scale (the unit of measurement) is the culprit is correct. But the idea of diminishing the size of the unit of measurement of the INDEPENDENT variable is wrong. The effect of an increase of one million dollars is surely higher than the effect of an increase of just one dollar. Thus, expressing incomes in dollars instead of thousands of dollars would REDUCE the absolute value of the coefficient. What is needed is the reverse: INCREASE the unit of the independent (or reduce the unit of the dependent), to get coefficients with a higher absolute value.

Besides, as remarked in another message to this thread, the STANDARDIZED COEFFICIENTS (whose unit is the standard deviation of each variable) should not suffer this problem.

Hector

From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Johnny Amora
Sent: 06 March 2009 08:13
To: [hidden email]
Subject: Re: linear regression coefficient of zero

Hi Phil, Paul and Garry,

Garry is right that it is not zero, its something .00000something. Perhaps your independent is expressed in thousands or millions of dollars. If this is the case, then Paul's suggestions is the solution. Re-scale your independent, say 100 millions instead of 100,000,000 by dividing 1,000, 000; or 100 thousands instead of 100, 000 by dividing 1000; etc.

Cheers,

Johnny

--- On Fri, 3/6/09, Garry Gelade <[hidden email]> wrote:

From: Garry Gelade <[hidden email].. uk >
Subject: Re: linear regression coefficient of zero
To: [hidden email]
Date: Friday, 6 March, 2009, 11:08 AM

When you say “the coefficient”, do you mean the standardized coefficient or the unstandardized coefficient? If it is the unstandardized coefficient that is zero, Paul’s method (re-scaling) will help you. In fact the coefficient is probably not zero, it is .00000something, which you can examine by increasing the number of decimal digits displayed.

If it is the standardized coefficient that is zero, you have a real problem!

Garry Gelade
Business Analytic.

From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Swank, Paul R
Sent: 05 March 2009 17:25
To: [hidden email]
Subject: Re: linear regression coefficient of zero

Have you tried rescaling it, say by diving the variable by 100?

Dr. Paul R. Swank,
Professor and Director of Research
Children's Learning Institute
University of Texas Health Science Center-Houston

From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Phil Schubert
Sent: Thursday, March 05, 2009 11:08 AM
To: [hidden email]
Subject: linear regression coefficient of zero

i'm an inexperienced user trying to complete a dissertation. in a recent linear regression model, the SPSS output indicated that one of my independent variables had a coefficient value of zero, despite showing to be statistically significant. The independent variable was pulled in from from excel and was originally in dollar form with the $ symbol and decimals. From the data editor, i changed the variable to a numeric, scale variable. I also attempted a second regression model, using only this one independent variable that i had trouble with. Again it showed to be significant, but the coefficient value was zero. I assume this has something to do with the way i have the data entered/coded. Any ideas on how to fix this would be appreciated.
--
Phil Schubert
Abilene Christian University
Executive Vice President
ACU Box 29120
Abilene , Texas 79699-9120
325/674-2795

__________ NOD32 3911 (20090305) Information __________

This message was checked by NOD32 antivirus system.
http://www.eset.com

Importing contacts has never been easier.
Bring your friends over to Yahoo! Mail today!

Interested in growing your business? Find out how with Yahoo! Search Marketing!

Martin P. Holt-2

Re: linear regression coefficient of zero

Yes. It's the direction of the change that Hector was referring to.

BW,

Martin

----- Original Message -----

From: [hidden email]

To: [hidden email]

Sent: Friday, March 06, 2009 4:56 PM

Subject: Re: linear regression coefficient of zero

The suggestion was just the same as using 1 kilogram, instead of 1000grams. I dont think such transformation in the indepedent is wrong. I dont think this is wrong: 1,000,000 USD= 1Million USD

--- On Sat, 3/7/09, Hector Maletta <[hidden email]> wrote:

From: Hector Maletta <[hidden email]>
Subject: Re: linear regression coefficient of zero
To: [hidden email]
Date: Saturday, 7 March, 2009, 12:06 AM

The general idea that the scale (the unit of measurement) is the culprit is correct. But the idea of diminishing the size of the unit of measurement of the INDEPENDENT variable is wrong. The effect of an increase of one million dollars is surely higher than the effect of an increase of just one dollar. Thus, expressing incomes in dollars instead of thousands of dollars would REDUCE the absolute value of the coefficient. What is needed is the reverse: INCREASE the unit of the independent (or reduce the unit of the dependent), to get coefficients with a higher absolute value.

Besides, as remarked in another message to this thread, the STANDARDIZED COEFFICIENTS (whose unit is the standard deviation of each variable) should not suffer this problem.

Hector

From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Johnny Amora
Sent: 06 March 2009 08:13
To: [hidden email]
Subject: Re: linear regression coefficient of zero

Hi Phil, Paul and Garry,

Garry is right that it is not zero, its something .00000something. Perhaps your independent is expressed in thousands or millions of dollars. If this is the case, then Paul's suggestions is the solution. Re-scale your independent, say 100 millions instead of 100,000,000 by dividing 1,000, 000; or 100 thousands instead of 100, 000 by dividing 1000; etc.

Cheers,

Johnny

--- On Fri, 3/6/09, Garry Gelade <[hidden email]> wrote:

From: Garry Gelade <[hidden email].. uk >
Subject: Re: linear regression coefficient of zero
To: [hidden email]
Date: Friday, 6 March, 2009, 11:08 AM

When you say “the coefficient”, do you mean the standardized coefficient or the unstandardized coefficient? If it is the unstandardized coefficient that is zero, Paul’s method (re-scaling) will help you. In fact the coefficient is probably not zero, it is .00000something, which you can examine by increasing the number of decimal digits displayed.

If it is the standardized coefficient that is zero, you have a real problem!

Garry Gelade

Business Analytic.

From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Swank, Paul R
Sent: 05 March 2009 17:25
To: [hidden email]
Subject: Re: linear regression coefficient of zero

Have you tried rescaling it, say by diving the variable by 100?

Dr. Paul R. Swank,

Professor and Director of Research

Children's Learning Institute

University of Texas Health Science Center-Houston

From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Phil Schubert
Sent: Thursday, March 05, 2009 11:08 AM
To: [hidden email]
Subject: linear regression coefficient of zero

i'm an inexperienced user trying to complete a dissertation. in a recent linear regression model, the SPSS output indicated that one of my independent variables had a coefficient value of zero, despite showing to be statistically significant. The independent variable was pulled in from from excel and was originally in dollar form with the $ symbol and decimals. From the data editor, i changed the variable to a numeric, scale variable. I also attempted a second regression model, using only this one independent variable that i had trouble with. Again it showed to be significant, but the coefficient value was zero. I assume this has something to do with the way i have the data entered/coded. Any ideas on how to fix this would be appreciated.
--
Phil Schubert
Abilene Christian University
Executive Vice President
ACU Box 29120
Abilene , Texas 79699-9120
325/674-2795

__________ NOD32 3911 (20090305) Information __________

This message was checked by NOD32 antivirus system.
http://www.eset.com

Importing contacts has never been easier.
Bring your friends over to Yahoo! Mail today!

Interested in growing your business? Find out how with Yahoo! Search Marketing!