SPSSX Discussion

Regression Analysis

Classic

List

Threaded

4 messages Options

Matt Donley

Regression Analysis

Hello all,

My name is Matt, I am a third year psychology student at Monash University
in Melbourne, Australia.
My group is conducting a research project into the relationship between
alcohol consumption and sex-related alcohol expectancies, with gender and
age as moderating factors. In order to obtain results, we are running a
hierarchical multiple regression using SPSS v15.0 but have encountered a
problem.
After centering relevant variables including age and recoding gender using
dummy variables (0,1) and performing interaction terms, we ran the
regression analysis.
First level: sexual expectancies
Second level: age and gender
Third level: interaction terms
We found significant results for enhancement (a sexual expectancy) and
genderXenhancement. We are now calculating an interaction graph and
require unstandardised correlation coefficients but for some reason SPSS
has excluded gender.

Can anyone please suggest possible solutions as to how we would go about
getting this data from the regression analysis? We have tried recoding the
gender data as suggested by our tutor but we still do not get the data we
need to produce the interaction graph. Any help would be greatly
appreciated as we are really sturggling and running out of time before we
have to hand it in.

Thank-you in advance,

Matt

Björn Türoque

Re: Regression Analysis

Matt,

This may sound like a dumb question, but did you happen to make two gender
variables one for male (values 1 and 0) and one for female (values 1 and
))? Putting both of them in will cause problems, but I don't know if it will
cause the problem you are talking about. For your regression you should only
use one variable for gender.

I only bring this possiblity up because one very simple mistake I often see
when people first run regression with recoded dummy variables is that they
place all of the dummy variables into the equation, when you need to leave
at least one out.

On 10/2/07, Matt Donley <[hidden email]> wrote:

>
> Hello all,
>
> My name is Matt, I am a third year psychology student at Monash University
> in Melbourne, Australia.
> My group is conducting a research project into the relationship between
> alcohol consumption and sex-related alcohol expectancies, with gender and
> age as moderating factors. In order to obtain results, we are running a
> hierarchical multiple regression using SPSS v15.0 but have encountered a
> problem.
> After centering relevant variables including age and recoding gender using
> dummy variables (0,1) and performing interaction terms, we ran the
> regression analysis.
> First level: sexual expectancies
> Second level: age and gender
> Third level: interaction terms
> We found significant results for enhancement (a sexual expectancy) and
> genderXenhancement. We are now calculating an interaction graph and
> require unstandardised correlation coefficients but for some reason SPSS
> has excluded gender.
>
> Can anyone please suggest possible solutions as to how we would go about
> getting this data from the regression analysis? We have tried recoding the
> gender data as suggested by our tutor but we still do not get the data we
> need to produce the interaction graph. Any help would be greatly
> appreciated as we are really sturggling and running out of time before we
> have to hand it in.
>
> Thank-you in advance,
>
> Matt
>

--
Björn Türoque
375 Hudson Street
New York, NY 10014-3657
212-366-2000

Some people are just born to rock!

Hector Maletta

Re: Regression Analysis

With regard to Bjorn's comment:
If gender is coded into two dummies the entire regression exercise
would be impossible due to perfect collinearity. Only one dummy should be
used, such as Male=1 and Female=0.
If regression is run in a stepwise fashion, SPSS may exclude any
variable (e.g. gender) because it is not a good predictor and thus fails to
fulfil the criteria for inclusion. Likewise, even if the variable is
included in the equation, an interaction involving that variable may be
excluded.

Now, returning to the original Matt's message, apparently gender
was NOT excluded from the equation, even gender interaction terms. The
trouble is with the GRAPH involving that interaction. On that particular
issue I cannot give any light. My only comment is about the talk about
standardized and unstandardized coefficients. Unstandardized coefficients
would give the effect of being male (gender=1) relative to being female
(gender=0) if that is the way gender was coded (or the reverse if the
opposite is the case). Standardized coefficients would represent the effect
of each gender RELATIVE TO THE "AVERAGE GENDER", MEASURED IN STANDARD
DEVIATIONS. The average gender is the proportion of males in the sample, p,
and the standard deviation is the square root of p(1-p). This is a convolute
and totally dumb way of representing the effect of gender, in my humble and
possibly ignorant opinion. I would stick with the plain unstandardized
coefficients, telling me how the effect changes when people is male compared
with the effect when they are female.

Hector

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Björn Türoque
Sent: 03 October 2007 10:31
To: [hidden email]
Subject: Re: Regression Analysis

Matt,

This may sound like a dumb question, but did you happen to make two
gender
variables one for male (values 1 and 0) and one for female (values
1 and
))? Putting both of them in will cause problems, but I don't know
if it will
cause the problem you are talking about. For your regression you
should only
use one variable for gender.

I only bring this possiblity up because one very simple mistake I
often see
when people first run regression with recoded dummy variables is
that they
place all of the dummy variables into the equation, when you need
to leave
at least one out.

On 10/2/07, Matt Donley <[hidden email]> wrote:
>
> Hello all,
>
> My name is Matt, I am a third year psychology student at Monash
University
> in Melbourne, Australia.
> My group is conducting a research project into the relationship
between
> alcohol consumption and sex-related alcohol expectancies, with
gender and
> age as moderating factors. In order to obtain results, we are
running a
> hierarchical multiple regression using SPSS v15.0 but have
encountered a
> problem.
> After centering relevant variables including age and recoding
gender using
> dummy variables (0,1) and performing interaction terms, we ran
the
> regression analysis.
> First level: sexual expectancies
> Second level: age and gender
> Third level: interaction terms
> We found significant results for enhancement (a sexual
expectancy) and
> genderXenhancement. We are now calculating an interaction graph
and
> require unstandardised correlation coefficients but for some
reason SPSS
> has excluded gender.
>
> Can anyone please suggest possible solutions as to how we would
go about
> getting this data from the regression analysis? We have tried
recoding the
> gender data as suggested by our tutor but we still do not get the
data we
> need to produce the interaction graph. Any help would be greatly
> appreciated as we are really sturggling and running out of time
before we
> have to hand it in.
>
> Thank-you in advance,
>
> Matt
>

--
Björn Türoque
375 Hudson Street
New York, NY 10014-3657
212-366-2000

Some people are just born to rock!

Ornelas, Fermin-2

Re: Regression Analysis

In reply to this post by Björn Türoque

Whenever you have dummy variables you will always have one dummy less fit into the model. As Hector already pointed out you want to avoid perfect correlation between two predictors. The effect on the missing dummy is collapsed into the constant coefficient.

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Björn Türoque
Sent: Wednesday, October 03, 2007 6:31 AM
To: [hidden email]
Subject: Re: Regression Analysis

Matt,

This may sound like a dumb question, but did you happen to make two gender
variables one for male (values 1 and 0) and one for female (values 1 and
))? Putting both of them in will cause problems, but I don't know if it will
cause the problem you are talking about. For your regression you should only
use one variable for gender.

I only bring this possiblity up because one very simple mistake I often see
when people first run regression with recoded dummy variables is that they
place all of the dummy variables into the equation, when you need to leave
at least one out.

On 10/2/07, Matt Donley <[hidden email]> wrote:

--
Björn Türoque
375 Hudson Street
New York, NY 10014-3657
212-366-2000

Some people are just born to rock!

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.