Re: dummy variable coding

[Yampolskaya, Svetlana]

Dear List,

I have a categorical variable that represents four reasons for discharge
from foster care: (a) reunification with parents, (b) placement with
relatives, c) adoption, and (d) other reasons not related to achieving
permanency. I tried to create four variables with the following coding:
1 - reunification and 0 - other reasons; 1 - placements with relatives
and 0 - other reasons, etc. However, whether I use "reunification" as a
reference category (i.e., when I don't include it in the multivariate
model) or "other reasons not related to permanency" as a reference
category I get different results.

Should I try contrast variable coding? If yes, can you please explain
how to do it and how it differs from regular dummy coding.

Any help will be appreciated,

Lana

[Dominic Lusinchi]

Lana,

You don't need 4 variables: only 3. If you use "reunification" as your
reference category, it will be coded 0 on all other 3 variables. The results
of your analysis will compare each indicator (dummy) variable to your
reference category.

Hope this is clear.

Good luck.

Dominic Lusinchi
Statistician
Far West Research
Statistical Consulting
San Francisco, California
415-664-3032
www.farwestresearch.com
-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Yampolskaya, Svetlana
Sent: Wednesday, August 16, 2006 12:21 PM
To: [hidden email]
Subject: Re: dummy variable coding

Dear List,

I have a categorical variable that represents four reasons for discharge
from foster care: (a) reunification with parents, (b) placement with
relatives, c) adoption, and (d) other reasons not related to achieving
permanency. I tried to create four variables with the following coding:
1 - reunification and 0 - other reasons; 1 - placements with relatives
and 0 - other reasons, etc. However, whether I use "reunification" as a
reference category (i.e., when I don't include it in the multivariate
model) or "other reasons not related to permanency" as a reference
category I get different results.

Should I try contrast variable coding? If yes, can you please explain
how to do it and how it differs from regular dummy coding.

Any help will be appreciated,

Lana

[Marks, Jim]

Lana:

The coefficients on the included variables give the influence of the
inlcuded variable compared to the reference category.

If the reference category changes, the comparison changes and the
influence relative to the (new) reference category will be different.

HTH

--jim


-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Yampolskaya, Svetlana
Sent: Wednesday, August 16, 2006 2:21 PM
To: [hidden email]
Subject: Re: dummy variable coding

Dear List,

I have a categorical variable that represents four reasons for discharge
from foster care: (a) reunification with parents, (b) placement with
relatives, c) adoption, and (d) other reasons not related to achieving
permanency. I tried to create four variables with the following coding:
1 - reunification and 0 - other reasons; 1 - placements with relatives
and 0 - other reasons, etc. However, whether I use "reunification" as a
reference category (i.e., when I don't include it in the multivariate
model) or "other reasons not related to permanency" as a reference
category I get different results.

Should I try contrast variable coding? If yes, can you please explain
how to do it and how it differs from regular dummy coding.

Any help will be appreciated,

Lana

[Maguin, Eugene]

All,

I'm now able to get plots. Now my question is this. In some of my glm models
I have a moderator and where the moderator main effect or or interactions
involving it are significant, I want plots within each moderator value. One
way to get them is to use a Split files command followed by the Igraph
command. However, for some unknown and stupid reason, spss puts the plots
side by side, which makes them pretty small. The Separate and Layer keywords
that control control how tabled data are presented on the page have no
effect on plots. Obviously, Temporary, Select if is an option. However, is
there another option that yields more compact syntax?

Thanks, Gene Maguin

[Beadle, ViAnn]

The graph command produces separate plots rather than paneled plots.

________________________________

From: SPSSX(r) Discussion on behalf of Gene Maguin
Sent: Wed 8/16/2006 3:36 PM
To: [hidden email]
Subject: Another plotting question



All,

I'm now able to get plots. Now my question is this. In some of my glm models
I have a moderator and where the moderator main effect or or interactions
involving it are significant, I want plots within each moderator value. One
way to get them is to use a Split files command followed by the Igraph
command. However, for some unknown and stupid reason, spss puts the plots
side by side, which makes them pretty small. The Separate and Layer keywords
that control control how tabled data are presented on the page have no
effect on plots. Obviously, Temporary, Select if is an option. However, is
there another option that yields more compact syntax?

Thanks, Gene Maguin

[Hector Maletta]

Some codification schemes allow contrasting each category not with another
category but, for instance, with the average effect of all categories. In
this case you do not have to omit one category. This and other types of
contrast are available in some SPSS procedures dealing with categorical
variables.
Hector

-----Mensaje original-----
De: SPSSX(r) Discussion [mailto:[hidden email]] En nombre de
Dominic Lusinchi
Enviado el: Wednesday, August 16, 2006 4:29 PM
Para: [hidden email]
Asunto: Re: dummy variable coding

Lana,

You don't need 4 variables: only 3. If you use "reunification" as your
reference category, it will be coded 0 on all other 3 variables. The results
of your analysis will compare each indicator (dummy) variable to your
reference category.

Hope this is clear.

Good luck.

Dominic Lusinchi
Statistician
Far West Research
Statistical Consulting
San Francisco, California
415-664-3032
www.farwestresearch.com
-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Yampolskaya, Svetlana
Sent: Wednesday, August 16, 2006 12:21 PM
To: [hidden email]
Subject: Re: dummy variable coding

Dear List,

I have a categorical variable that represents four reasons for discharge
from foster care: (a) reunification with parents, (b) placement with
relatives, c) adoption, and (d) other reasons not related to achieving
permanency. I tried to create four variables with the following coding:
1 - reunification and 0 - other reasons; 1 - placements with relatives
and 0 - other reasons, etc. However, whether I use "reunification" as a
reference category (i.e., when I don't include it in the multivariate
model) or "other reasons not related to permanency" as a reference
category I get different results.

Should I try contrast variable coding? If yes, can you please explain
how to do it and how it differs from regular dummy coding.

Any help will be appreciated,

Lana