three way interactions categorical predictor and two continuious moderators

[ann-2]

Hi everybody,

i have a question i am kinda stuck with.

i want to run a regression model with a categorical predictor (3
categories) predicting a continuous outcomes but at the time also include
two continuous moderators.

i have dummy coded my categorical independent variable and centered the
continuous moderators. For the investigation of potential moderation i have
to multiply the first moderator (m1) with the two dummy variables(d1 and
d2), then i multipy the second moderator (m2) with the two dummy coded
variables. and how do i do to investigate the three way interaction.

so far:
d1*m1
d2*m1
d1*m2
d2*m2

so do i just multiply all together for the three way interaction?--> (d1*m1)
*(d2*m1)*(d1*m2)*(d1*m2)

or is the whole procedure better done using ANOVA?

i am kinda lost here, so for some advice on how to proceed as well as some
reading hints I would be grateful

thanks

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

[Maguin, Eugene]

Ann,

It doesn't sound like you are interested in the triple interaction (iv with
m1 with m2). If so, there in no need to include it. The regression model
would have terms for d1, d2, m1, m2, d1m1, d1m2, d2m1, d2m2, plus the
constant. You could certainly do this with anova but you will have to
specify the effects on the design statement so that you don't get the triple
interaction term as well.

Gene Maguin





-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of ann
Sent: Monday, August 16, 2010 10:08 AM
To: [hidden email]
Subject: three way interactions categorical predictor and two continuious
moderators

Hi everybody,

i have a question i am kinda stuck with.

i want to run a regression model with a categorical predictor (3
categories) predicting a continuous outcomes but at the time also include
two continuous moderators.

i have dummy coded my categorical independent variable and centered the
continuous moderators. For the investigation of potential moderation i have
to multiply the first moderator (m1) with the two dummy variables(d1 and
d2), then i multipy the second moderator (m2) with the two dummy coded
variables. and how do i do to investigate the three way interaction.

so far:
d1*m1
d2*m1
d1*m2
d2*m2

so do i just multiply all together for the three way interaction?--> (d1*m1)
*(d2*m1)*(d1*m2)*(d1*m2)

or is the whole procedure better done using ANOVA?

i am kinda lost here, so for some advice on how to proceed as well as some
reading hints I would be grateful

thanks

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

[Dale Glaser]

Hi all...I did use the missing values module when first released (I think it was V10 or 9?) but since then have tended to use the MI option (EM algorithim with MCMC) in the Prelis preprocessor or the missing data default in Mplus (FIML), though with Mplus6.0 there are some new additions for missing data (e.g, MI using Bayesian analysis); moreover, it was many years since I used NORM (it was DOS based back then)..............so I would be most appreciative to get your opinions on the SPSS missing data module, and if it would add much beyond the PRELIS and or Mplus options (or should I just download NORM for MI)?  I ask this insofar I have read some of the software comparisons in American Statistician for missing data, and, at least one I read a few years ago, was not too encouraging about the EM option in SPSS...........thank you very much for your feedback......Dale

Dale Glaser, Ph.D.
Principal--Glaser Consulting
Lecturer/Adjunct Faculty--SDSU/USD/Alliant
Past-President, San Diego Chapter of
American Statistical Association
3115 4th Avenue
San Diego, CA 92103
phone: [phone]
fax: [phone]
email: [hidden email]
website: www.glaserconsult.com

 

[ann-2]

hi

thanks for ur reply

yes i am interested in the three-way interaction, that is why i run into
those problems where i dont know how this multiplicatory term, expressing
the three-way interaction, is supposed to look.

anybody an idea?

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

[Bruce Weaver]

ann-2 wrote

hi

thanks for ur reply

yes i am interested in the three-way interaction, that is why i run into
those problems where i dont know how this multiplicatory term, expressing
the three-way interaction, is supposed to look.

anybody an idea?

Y = b0 +
     b1*d1 +
     b2*d2 +
     b3*m1 +
     b4*m2 +
     b5*d1*m1 +
     b6*d2*m1 +
     b7*d1*m2 +
     b8*d2*m2 +
     b9*d1*m1*m2 +
     b10*d2*m1*m2 +
     error

[Ryan]

Hi Bruce,

You're missing the two-way interaction between m1 and m2.

Ryan

On Tue, Aug 17, 2010 at 9:48 AM, Bruce Weaver <[hidden email]> wrote:

> ann-2 wrote:
>>
>> hi
>>
>> thanks for ur reply
>>
>> yes i am interested in the three-way interaction, that is why i run into
>> those problems where i dont know how this multiplicatory term, expressing
>> the three-way interaction, is supposed to look.
>>
>> anybody an idea?
>>
>>
>
>
> Y = b0 +
>     b1*d1 +
>     b2*d2 +
>     b3*m1 +
>     b4*m2 +
>     b5*d1*m1 +
>     b6*d2*m1 +
>     b7*d1*m2 +
>     b8*d2*m2 +
>     b9*d1*m1*m2 +
>     b10*d2*m1*m2 +
>     error
>
>
>
> -----
> --
> Bruce Weaver
> [hidden email]
> http://sites.google.com/a/lakeheadu.ca/bweaver/
>
> "When all else fails, RTFM."
>
> NOTE: My Hotmail account is not monitored regularly.
> To send me an e-mail, please use the address shown above.
>
> --
> View this message in context: http://spssx-discussion.1045642.n5.nabble.com/three-way-interactions-categorical-predictor-and-two-continuious-moderators-tp2636899p2638162.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

[Ryan]

Scratch the last comment. I see the OP is not interested in the
four-way interaction. Bruce's equation looks correct. -Ryan

On Tue, Aug 17, 2010 at 9:52 AM, R B <[hidden email]> wrote:

> Hi Bruce,
>
> You're missing the two-way interaction between m1 and m2.
>
> Ryan
>
> On Tue, Aug 17, 2010 at 9:48 AM, Bruce Weaver <[hidden email]> wrote:
>> ann-2 wrote:
>>>
>>> hi
>>>
>>> thanks for ur reply
>>>
>>> yes i am interested in the three-way interaction, that is why i run into
>>> those problems where i dont know how this multiplicatory term, expressing
>>> the three-way interaction, is supposed to look.
>>>
>>> anybody an idea?
>>>
>>>
>>
>>
>> Y = b0 +
>>     b1*d1 +
>>     b2*d2 +
>>     b3*m1 +
>>     b4*m2 +
>>     b5*d1*m1 +
>>     b6*d2*m1 +
>>     b7*d1*m2 +
>>     b8*d2*m2 +
>>     b9*d1*m1*m2 +
>>     b10*d2*m1*m2 +
>>     error
>>
>>
>>
>> -----
>> --
>> Bruce Weaver
>> [hidden email]
>> http://sites.google.com/a/lakeheadu.ca/bweaver/
>>
>> "When all else fails, RTFM."
>>
>> NOTE: My Hotmail account is not monitored regularly.
>> To send me an e-mail, please use the address shown above.
>>
>> --
>> View this message in context: http://spssx-discussion.1045642.n5.nabble.com/three-way-interactions-categorical-predictor-and-two-continuious-moderators-tp2636899p2638162.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

[Bruce Weaver]

R B wrote

Scratch the last comment. I see the OP is not interested in the
four-way interaction. Bruce's equation looks correct. -Ryan

On Tue, Aug 17, 2010 at 9:52 AM, R B <[e-mail]> wrote:
> Hi Bruce,
>
> You're missing the two-way interaction between m1 and m2.
>
> Ryan
>

I think you were right the first time, Ryan.  If the 3-way interaction is in the model, all of the component two-way interactions should be there too.  But I forgot to include m1*m2.  So it should be:

Y = b0 +         { the constant }

     b1*d1 +     { main effects }
     b2*d2 +
     b3*m1 +
     b4*m2 +

     b5*d1*m1 +    { two-way interactions }
     b6*d2*m1 +
     b7*d1*m2 +
     b8*d2*m2 +
     b9*m1*m2 +     { this term was missing from my first post }

     b10*d1*m1*m2 +   { the three-way interaction }
     b11*d2*m1*m2 +

     error

[Ryan]

Yes, all two-way interactions need to be included in the presence of a
three-way interaction. Not sure why I second-guessed myself. Too much
multitasking, I tell ya. :)

On Tue, Aug 17, 2010 at 10:28 AM, Bruce Weaver <[hidden email]> wrote:

> R B wrote:
>>
>> Scratch the last comment. I see the OP is not interested in the
>> four-way interaction. Bruce's equation looks correct. -Ryan
>>
>> On Tue, Aug 17, 2010 at 9:52 AM, R B <[hidden email]> wrote:
>>> Hi Bruce,
>>>
>>> You're missing the two-way interaction between m1 and m2.
>>>
>>> Ryan
>>>
>>
>
> I think you were right the first time, Ryan.  If the 3-way interaction is in
> the model, all of the component two-way interactions should be there too.
> But I forgot to include m1*m2.  So it should be:
>
> Y = b0 +         { the constant }
>
>     b1*d1 +     { main effects }
>     b2*d2 +
>     b3*m1 +
>     b4*m2 +
>
>     b5*d1*m1 +    { two-way interactions }
>     b6*d2*m1 +
>     b7*d1*m2 +
>     b8*d2*m2 +
>     b9*m1*m2 +     { this term was missing from my first post }
>
>     b10*d1*m1*m2 +   { the three-way interaction }
>     b11*d2*m1*m2 +
>
>     error
>
>
>
> -----
> --
> Bruce Weaver
> [hidden email]
> http://sites.google.com/a/lakeheadu.ca/bweaver/
>
> "When all else fails, RTFM."
>
> NOTE: My Hotmail account is not monitored regularly.
> To send me an e-mail, please use the address shown above.
>
> --
> View this message in context: http://spssx-discussion.1045642.n5.nabble.com/three-way-interactions-categorical-predictor-and-two-continuious-moderators-tp2636899p2638218.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

[Jon K Peck]

The SPSS Developer Central site is migrating and morphing to the IBM myDeveloperWorks site, where it will be integrated with other IBM pages.

DevCentral has provided programmability plugins, supplementary downloads, articles, forums, and a blog.  These will all be part of the SPSS group in myDW.  We are migrating most of the old content apart from a few obsolete items.  The new site will give us access to better tools for the pages, including wikis and feeds, so it should be more convenient and useful, better organized, more searchable, and easier to evolve forward.

The complete site will become visible soon, and you will be able to join the SPSS group and see all of the facilities.

The first step has happened today.  The DevCentral forums have been replaced with myDW forums.  The existing content has been migrated, but all the posts are by "Guest".  If you go to the old forums, you will be redirected to the new site.  You can also go there directly via this url:

http://www.ibm.com/developerworks/forums/category.jspa?categoryID=302

Once the SPSS group is public, you will be able to click a link on the group front page to go there.

Anyone can read the posts, but to do a post, you must be registered on myDeveloperWorks.  Please visit and create your new registration.  (We could not migrate the old registration data, because IBM requires slightly different data than the SPSS site does.)

One of the benefits of the new forum software is that you can watch threads, so if you post a question, you can get email when it is answered.

All of this is new, so things will likely change as we develop more experience with it.

Please make yourselves at home in our new house.

Regards,

Jon Peck
SPSS, an IBM Company
[hidden email]
312-651-3435