Controlling for Race with SPSS 20

[Justin Blehar]

Hello All,

Not sure how much detail is needed so I'll give you a quick overall but I'm trying to control for race and am unsure how to best go about this. I know that I can run a partial correlation and control for race using the menu but is this really controlling for race? If not is there a better way? How would I do this for a t-test?

This is a cross sectional design looking at cognition and smoking in a psychiatric population. There are six groups I'm looking at; Never Smokers, Former Smokers, Nonsmokers (includes both never smokers and former smokers), Heavy Smokers, Light Smokers, and Smokers (includes heavy and light smokers). I have 36 scale variables that I want to compare between each of these groups. When I break out the groups by race (just looking at box plots and mean comparisons) there are clearly some large race effects (e.g. parental education, level of functioning, IQ, etc...). I'd like to be able to correct for this in each analysis. I'm running both correlations and t-tests (maybe this isn't the best process?).

If I run a partial correlation and control for race is this really controlling for race?

When running the t-tests how do I control for race?

Any help would be greatly appreciated.

V/R

Justin  

[Mike]

If you control for race in whatever manner, to what population
would your conclusions apply to?  And by population I mean
humans, not a mathematical distribution.

-Mike Palij
New York University
[hidden email]


On Sat, Sep 29, 2012 at 6:18 PM, Justin Blehar <[hidden email]> wrote:

> Hello All,
>
> Not sure how much detail is needed so I'll give you a quick overall but I'm
> trying to control for race and am unsure how to best go about this. I know
> that I can run a partial correlation and control for race using the menu but
> is this really controlling for race? If not is there a better way? How would
> I do this for a t-test?
>
> This is a cross sectional design looking at cognition and smoking in a
> psychiatric population. There are six groups I'm looking at; Never Smokers,
> Former Smokers, Nonsmokers (includes both never smokers and former smokers),
> Heavy Smokers, Light Smokers, and Smokers (includes heavy and light
> smokers). I have 36 scale variables that I want to compare between each of
> these groups. When I break out the groups by race (just looking at box plots
> and mean comparisons) there are clearly some large race effects (e.g.
> parental education, level of functioning, IQ, etc...). I'd like to be able
> to correct for this in each analysis. I'm running both correlations and
> t-tests (maybe this isn't the best process?).
>
> If I run a partial correlation and control for race is this really
> controlling for race?
>
> When running the t-tests how do I control for race?
>
> Any help would be greatly appreciated.
>
> V/R
>
> Justin

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

[Justin Blehar]

The population is outpatient individuals suffering from schizophrenia. Most of the research I found either does not list race or only has Caucasians in their sample. I have a large enough N to run only Caucasians but this limits how well the results can be generalized. I'm trying to avoid this if possible.

Total Sample
Caucasian - N = 76
African American- N = 44

Groups
Never Smokers: Caucasian N = 15 African American N = 23
Former Smokers: Caucasian N = 29 African American N = 2
Nonsmokers: Caucasian N = 44 African American N = 25
Heavy Smokers: Caucasian N = 9 African American N = 5
Light Smokers: Caucasian N = 23 African American N =14
Current Smokers: Caucasian N = 32 African American N = 19

Thanks for your reply :)

V/R
Justin
________________________________________
From: SPSSX(r) Discussion [[hidden email]] on behalf of Michael Palij [[hidden email]]
Sent: Saturday, September 29, 2012 8:11 PM
To: [hidden email]
Subject: Re: Controlling for Race with SPSS 20

If you control for race in whatever manner, to what population
would your conclusions apply to?  And by population I mean
humans, not a mathematical distribution.

-Mike Palij
New York University
[hidden email]


On Sat, Sep 29, 2012 at 6:18 PM, Justin Blehar <[hidden email]> wrote:

> Hello All,
>
> Not sure how much detail is needed so I'll give you a quick overall but I'm
> trying to control for race and am unsure how to best go about this. I know
> that I can run a partial correlation and control for race using the menu but
> is this really controlling for race? If not is there a better way? How would
> I do this for a t-test?
>
> This is a cross sectional design looking at cognition and smoking in a
> psychiatric population. There are six groups I'm looking at; Never Smokers,
> Former Smokers, Nonsmokers (includes both never smokers and former smokers),
> Heavy Smokers, Light Smokers, and Smokers (includes heavy and light
> smokers). I have 36 scale variables that I want to compare between each of
> these groups. When I break out the groups by race (just looking at box plots
> and mean comparisons) there are clearly some large race effects (e.g.
> parental education, level of functioning, IQ, etc...). I'd like to be able
> to correct for this in each analysis. I'm running both correlations and
> t-tests (maybe this isn't the best process?).
>
> If I run a partial correlation and control for race is this really
> controlling for race?
>
> When running the t-tests how do I control for race?
>
> Any help would be greatly appreciated.
>
> V/R
>
> Justin

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

[Mike]

The simple answer to your question is that you use race as a
grouping variable and you test for the whether the effect is the
same across levels of race, that is, a nonsignificant two-way
interaction -- assuming you have sufficient statistical power to
detect such an interaction.  If you have a significant two-way
interaction, that is, effects are different for the two groups,
then you can not come up with a result that applies
to all persons independent of race -- the effect is dependent
upon the race of the person.  If the effect is nonsignificant
with adequate power, then the effect holds for the races
represented in the design.

Given the sample sizes you list below, I think descriptive
analyses are more appropriate than inferential analyses but
what the hell, who knows, maybe N=2 does provide sufficient
information for statistical testing.  Knock yourself out.

-Mike Palij
New York University
[hidden email]


On Sat, Sep 29, 2012 at 8:32 PM, Justin Blehar <[hidden email]> wrote:

> The population is outpatient individuals suffering from schizophrenia. Most of the research I found either does not list race or only has Caucasians in their sample. I have a large enough N to run only Caucasians but this limits how well the results can be generalized. I'm trying to avoid this if possible.
>
> Total Sample
> Caucasian - N = 76
> African American- N = 44
>
> Groups
> Never Smokers: Caucasian N = 15 African American N = 23
> Former Smokers: Caucasian N = 29 African American N = 2
> Nonsmokers: Caucasian N = 44 African American N = 25
> Heavy Smokers: Caucasian N = 9 African American N = 5
> Light Smokers: Caucasian N = 23 African American N =14
> Current Smokers: Caucasian N = 32 African American N = 19
>
> Thanks for your reply :)
>
> V/R
> Justin
> ________________________________________
> From: SPSSX(r) Discussion [[hidden email]] on behalf of Michael Palij [[hidden email]]
> Sent: Saturday, September 29, 2012 8:11 PM
> To: [hidden email]
> Subject: Re: Controlling for Race with SPSS 20
>
> If you control for race in whatever manner, to what population
> would your conclusions apply to?  And by population I mean
> humans, not a mathematical distribution.
>
> -Mike Palij
> New York University
> [hidden email]
>
>
> On Sat, Sep 29, 2012 at 6:18 PM, Justin Blehar <[hidden email]> wrote:
>> Hello All,
>>
>> Not sure how much detail is needed so I'll give you a quick overall but I'm
>> trying to control for race and am unsure how to best go about this. I know
>> that I can run a partial correlation and control for race using the menu but
>> is this really controlling for race? If not is there a better way? How would
>> I do this for a t-test?
>>
>> This is a cross sectional design looking at cognition and smoking in a
>> psychiatric population. There are six groups I'm looking at; Never Smokers,
>> Former Smokers, Nonsmokers (includes both never smokers and former smokers),
>> Heavy Smokers, Light Smokers, and Smokers (includes heavy and light
>> smokers). I have 36 scale variables that I want to compare between each of
>> these groups. When I break out the groups by race (just looking at box plots
>> and mean comparisons) there are clearly some large race effects (e.g.
>> parental education, level of functioning, IQ, etc...). I'd like to be able
>> to correct for this in each analysis. I'm running both correlations and
>> t-tests (maybe this isn't the best process?).
>>
>> If I run a partial correlation and control for race is this really
>> controlling for race?
>>
>> When running the t-tests how do I control for race?
>>
>> Any help would be greatly appreciated.
>>
>> V/R
>>
>> Justin
>
> =====================
> 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

=====================
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]

Given the sample sizes, I would forget about the distinctions between Never-Former and Light-Heavy.  That reduces it to a 2x2 (Smoker x Race) with decent numbers in every cell.
 
Nonsmokers: Caucasian N = 44 African American N = 25
Current Smokers: Caucasian N = 32 African American N = 19

HTH.

Mike Palij wrote

The simple answer to your question is that you use race as a
grouping variable and you test for the whether the effect is the
same across levels of race, that is, a nonsignificant two-way
interaction -- assuming you have sufficient statistical power to
detect such an interaction.  If you have a significant two-way
interaction, that is, effects are different for the two groups,
then you can not come up with a result that applies
to all persons independent of race -- the effect is dependent
upon the race of the person.  If the effect is nonsignificant
with adequate power, then the effect holds for the races
represented in the design.

Given the sample sizes you list below, I think descriptive
analyses are more appropriate than inferential analyses but
what the hell, who knows, maybe N=2 does provide sufficient
information for statistical testing.  Knock yourself out.

-Mike Palij
New York University
[hidden email]


On Sat, Sep 29, 2012 at 8:32 PM, Justin Blehar <[hidden email]> wrote:
> The population is outpatient individuals suffering from schizophrenia. Most of the research I found either does not list race or only has Caucasians in their sample. I have a large enough N to run only Caucasians but this limits how well the results can be generalized. I'm trying to avoid this if possible.
>
> Total Sample
> Caucasian - N = 76
> African American- N = 44
>
> Groups
> Never Smokers: Caucasian N = 15 African American N = 23
> Former Smokers: Caucasian N = 29 African American N = 2
> Nonsmokers: Caucasian N = 44 African American N = 25
> Heavy Smokers: Caucasian N = 9 African American N = 5
> Light Smokers: Caucasian N = 23 African American N =14
> Current Smokers: Caucasian N = 32 African American N = 19
>
> Thanks for your reply :)
>
> V/R
> Justin
> ________________________________________
> From: SPSSX(r) Discussion [[hidden email]] on behalf of Michael Palij [[hidden email]]
> Sent: Saturday, September 29, 2012 8:11 PM
> To: [hidden email]
> Subject: Re: Controlling for Race with SPSS 20
>
> If you control for race in whatever manner, to what population
> would your conclusions apply to?  And by population I mean
> humans, not a mathematical distribution.
>
> -Mike Palij
> New York University
> [hidden email]
>
>
> On Sat, Sep 29, 2012 at 6:18 PM, Justin Blehar <[hidden email]> wrote:
>> Hello All,
>>
>> Not sure how much detail is needed so I'll give you a quick overall but I'm
>> trying to control for race and am unsure how to best go about this. I know
>> that I can run a partial correlation and control for race using the menu but
>> is this really controlling for race? If not is there a better way? How would
>> I do this for a t-test?
>>
>> This is a cross sectional design looking at cognition and smoking in a
>> psychiatric population. There are six groups I'm looking at; Never Smokers,
>> Former Smokers, Nonsmokers (includes both never smokers and former smokers),
>> Heavy Smokers, Light Smokers, and Smokers (includes heavy and light
>> smokers). I have 36 scale variables that I want to compare between each of
>> these groups. When I break out the groups by race (just looking at box plots
>> and mean comparisons) there are clearly some large race effects (e.g.
>> parental education, level of functioning, IQ, etc...). I'd like to be able
>> to correct for this in each analysis. I'm running both correlations and
>> t-tests (maybe this isn't the best process?).
>>
>> If I run a partial correlation and control for race is this really
>> controlling for race?
>>
>> When running the t-tests how do I control for race?
>>
>> Any help would be greatly appreciated.
>>
>> V/R
>>
>> Justin
>
> =====================
> 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

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

[Justin Blehar]

Mike and Bruce,

Thanks so much for your input and it is helpful. I know power is an issue and collapsing the groups is a possibility but I'm concerned that I'd be missing something based on the descriptive stats I've already run (e.g. there is a large mean difference between the current smoker Caucasian group and working memory vs the current smoker African American group and working memory). The other option would be to just run the Caucasian sample. You've both given me some more to think about.

Thanks so much!

V/R
Justin
________________________________________
From: SPSSX(r) Discussion [[hidden email]] on behalf of Bruce Weaver [[hidden email]]
Sent: Saturday, September 29, 2012 11:10 PM
To: [hidden email]
Subject: Re: Controlling for Race with SPSS 20

Given the sample sizes, I would forget about the distinctions between
Never-Former and Light-Heavy.  That reduces it to a 2x2 (Smoker x Race) with
decent numbers in every cell.

Nonsmokers: Caucasian N = 44 African American N = 25
Current Smokers: Caucasian N = 32 African American N = 19

HTH.



Mike Palij wrote

> The simple answer to your question is that you use race as a
> grouping variable and you test for the whether the effect is the
> same across levels of race, that is, a nonsignificant two-way
> interaction -- assuming you have sufficient statistical power to
> detect such an interaction.  If you have a significant two-way
> interaction, that is, effects are different for the two groups,
> then you can not come up with a result that applies
> to all persons independent of race -- the effect is dependent
> upon the race of the person.  If the effect is nonsignificant
> with adequate power, then the effect holds for the races
> represented in the design.
>
> Given the sample sizes you list below, I think descriptive
> analyses are more appropriate than inferential analyses but
> what the hell, who knows, maybe N=2 does provide sufficient
> information for statistical testing.  Knock yourself out.
>
> -Mike Palij
> New York University

> mp26@

>
>
> On Sat, Sep 29, 2012 at 8:32 PM, Justin Blehar &lt;

> jnblehar@

> &gt; wrote:
>> The population is outpatient individuals suffering from schizophrenia.
>> Most of the research I found either does not list race or only has
>> Caucasians in their sample. I have a large enough N to run only
>> Caucasians but this limits how well the results can be generalized. I'm
>> trying to avoid this if possible.
>>
>> Total Sample
>> Caucasian - N = 76
>> African American- N = 44
>>
>> Groups
>> Never Smokers: Caucasian N = 15 African American N = 23
>> Former Smokers: Caucasian N = 29 African American N = 2
>> Nonsmokers: Caucasian N = 44 African American N = 25
>> Heavy Smokers: Caucasian N = 9 African American N = 5
>> Light Smokers: Caucasian N = 23 African American N =14
>> Current Smokers: Caucasian N = 32 African American N = 19
>>
>> Thanks for your reply :)
>>
>> V/R
>> Justin
>> ________________________________________
>> From: SPSSX(r) Discussion [

> SPSSX-L@.UGA

> ] on behalf of Michael Palij [

> mp26@

> ]
>> Sent: Saturday, September 29, 2012 8:11 PM
>> To:

> SPSSX-L@.UGA

>> Subject: Re: Controlling for Race with SPSS 20
>>
>> If you control for race in whatever manner, to what population
>> would your conclusions apply to?  And by population I mean
>> humans, not a mathematical distribution.
>>
>> -Mike Palij
>> New York University
>>

> mp26@

>>
>>
>> On Sat, Sep 29, 2012 at 6:18 PM, Justin Blehar &lt;

> jnblehar@

> &gt; wrote:
>>> Hello All,
>>>
>>> Not sure how much detail is needed so I'll give you a quick overall but
>>> I'm
>>> trying to control for race and am unsure how to best go about this. I
>>> know
>>> that I can run a partial correlation and control for race using the menu
>>> but
>>> is this really controlling for race? If not is there a better way? How
>>> would
>>> I do this for a t-test?
>>>
>>> This is a cross sectional design looking at cognition and smoking in a
>>> psychiatric population. There are six groups I'm looking at; Never
>>> Smokers,
>>> Former Smokers, Nonsmokers (includes both never smokers and former
>>> smokers),
>>> Heavy Smokers, Light Smokers, and Smokers (includes heavy and light
>>> smokers). I have 36 scale variables that I want to compare between each
>>> of
>>> these groups. When I break out the groups by race (just looking at box
>>> plots
>>> and mean comparisons) there are clearly some large race effects (e.g.
>>> parental education, level of functioning, IQ, etc...). I'd like to be
>>> able
>>> to correct for this in each analysis. I'm running both correlations and
>>> t-tests (maybe this isn't the best process?).
>>>
>>> If I run a partial correlation and control for race is this really
>>> controlling for race?
>>>
>>> When running the t-tests how do I control for race?
>>>
>>> Any help would be greatly appreciated.
>>>
>>> V/R
>>>
>>> Justin
>>
>> =====================
>> To manage your subscription to SPSSX-L, send a message to
>>

> LISTSERV@.UGA

>  (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
>>

> LISTSERV@.UGA

>  (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

> LISTSERV@.UGA

>  (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
[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/Controlling-for-Race-with-SPSS-20-tp5715386p5715390.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

[Justin Blehar]

Pooling some of the groups would work, I'm thinking a combination of descriptive and inferential stats is probably going to be the best course of action. Yes current smokers comprise both the heavy and light smokers so I am double counting cases. A 3 way design may work for this. My intent for the analysis is exploratory to gain a better understanding of how (if at all) smoking status impacts cognition in schizophrenia. The current literature is all over the place with mixed results and various designs. I'm hoping that by using the higher number of groups it can reduce confounds and paint a clearer picture (e.g. they're may be no relation when looking at current smokers and processing speed but a strong relation exists between heavy smokers and processing speed). Ideally I can look at each of these combinations between and within groups. All the while controlling for race.

Cognition is being assessed by: Vocabulary (VC), Matrix Reasoning(MR), Estimated IQ (from VC and MR scores). Ten other individual measures (TMT, BAC, HVLT, WMSlll, LNS, Mazes, BVMT, Fluency, MSCEIT, CPT) and raw scores are converted to T scores and then used to make up 6 cognitive domains  plus a total score (Processing Speed, Attention and Vigilance, Working Memory, Verbal Learning, Reasoning Speed, Social Cognition, and a Total Score for the Battery).

Thanks again and I appreciate your patience, I'm new to all of this and really enjoy learning from everyone!

V/R

Justin

---

From: Art Kendall [[hidden email]]
Sent: Sunday, September 30, 2012 8:31 AM
To: Justin Blehar
Cc: [hidden email]
Subject: Re: [SPSSX-L] Controlling for Race with SPSS 20

Since you have tiny Ns when grouping cases this way, perhaps you should pool some of the groups? Or use and additional IV? with a 3 way design
As is, it looks like you are double counting cases.
E.g., how do you distinguish heavy smokers from current smokers?

Can you put your cases into a 3 way design (3 * 2 *2), degree of exposure, race as IVs and cognition as the DV?
degree of exposure 3 levels: never, light, heavy
current status 2 levels: current/not
race 2 levels: Caucasian, African American.

How are you measuring cognition?
On 9/29/2012 8:32 PM, Justin Blehar wrote:

The population is outpatient individuals suffering from schizophrenia. Most of the research I found either does not list race or only has Caucasians in their sample. I have a large enough N to run only Caucasians but this limits how well the results can be generalized. I'm trying to avoid this if possible.

Total Sample
Caucasian - N = 76
African American- N = 44

Groups
Never Smokers: Caucasian N = 15 African American N = 23
Former Smokers: Caucasian N = 29 African American N = 2
Nonsmokers: Caucasian N = 44 African American N = 25
Heavy Smokers: Caucasian N = 9 African American N = 5
Light Smokers: Caucasian N = 23 African American N =14
Current Smokers: Caucasian N = 32 African American N = 19

Thanks for your reply :)

V/R
Justin
________________________________________
From: SPSSX(r) Discussion [[hidden email]] on behalf of Michael Palij [[hidden email]]
Sent: Saturday, September 29, 2012 8:11 PM
To: [hidden email]
Subject: Re: Controlling for Race with SPSS 20

If you control for race in whatever manner, to what population
would your conclusions apply to?  And by population I mean
humans, not a mathematical distribution.

-Mike Palij
New York University
[hidden email]


On Sat, Sep 29, 2012 at 6:18 PM, Justin Blehar [hidden email] wrote:

Hello All,

Not sure how much detail is needed so I'll give you a quick overall but I'm
trying to control for race and am unsure how to best go about this. I know
that I can run a partial correlation and control for race using the menu but
is this really controlling for race? If not is there a better way? How would
I do this for a t-test?

This is a cross sectional design looking at cognition and smoking in a
psychiatric population. There are six groups I'm looking at; Never Smokers,
Former Smokers, Nonsmokers (includes both never smokers and former smokers),
Heavy Smokers, Light Smokers, and Smokers (includes heavy and light
smokers). I have 36 scale variables that I want to compare between each of
these groups. When I break out the groups by race (just looking at box plots
and mean comparisons) there are clearly some large race effects (e.g.
parental education, level of functioning, IQ, etc...). I'd like to be able
to correct for this in each analysis. I'm running both correlations and
t-tests (maybe this isn't the best process?).

If I run a partial correlation and control for race is this really
controlling for race?

When running the t-tests how do I control for race?

Any help would be greatly appreciated.

V/R

Justin

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

[msherman]

Justin: A couple of random notes.

a.       Smoking status (current smoker vs non-current smoker). Including heavy and light smokers in your current smoker group is not double counting. You just have a mixed group of smokers.

b.      Smoking status cannot “impact” cognition given that you did not manipulate smoking status. All you have is an attribute variable which may well be confounded with some other variable or variables.

c.       I am not sure what you mean by reduce confounding by using a higher number of groups. Using simply current smoker vs. breaking current smoker into two groups of light smoker and heavy smoker is not an issue of confounding. Unless I am missing something.

d.      Looking at current smokers (Yes current or No) and processing speed is legitimate given that current smokers is a binary variable. Using only heavier smokers and processing speed does not lend itself to an analysis because you would only have one level of the variable. If on the other hand you meant to say that you had heaver smokers, light smokers and non-smokers than yes you could look for a relation. Or simply heavy smokers vs. nonsmokers-this would work also.

e.       In regard for controlling for race if you have a sufficient number of different races (White vs. nonwhite) why not use race a predictor variable and obtain information about the interaction between race and smoking status.

 

Martin F. Sherman, Ph.D.

Professor of Psychology

Director of  Masters Education in Psychology: Thesis Track

 

Loyola University Maryland

Department of Psychology

222 B Beatty Hall

4501 North Charles Street

Baltimore, MD 21210

 

410-617-2417

[hidden email]

 

 

 

From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Justin Blehar
Sent: Sunday, September 30, 2012 12:55 PM
To: [hidden email]
Subject: Re: Controlling for Race with SPSS 20

 

Pooling some of the groups would work, I'm thinking a combination of descriptive and inferential stats is probably going to be the best course of action. Yes current smokers comprise both the heavy and light smokers so I am double counting cases. A 3 way design may work for this. My intent for the analysis is exploratory to gain a better understanding of how (if at all) smoking status impacts cognition in schizophrenia. The current literature is all over the place with mixed results and various designs. I'm hoping that by using the higher number of groups it can reduce confounds and paint a clearer picture (e.g. they're may be no relation when looking at current smokers and processing speed but a strong relation exists between heavy smokers and processing speed). Ideally I can look at each of these combinations between and within groups. All the while controlling for race.

Cognition is being assessed by: Vocabulary (VC), Matrix Reasoning(MR), Estimated IQ (from VC and MR scores). Ten other individual measures (TMT, BAC, HVLT, WMSlll, LNS, Mazes, BVMT, Fluency, MSCEIT, CPT) and raw scores are converted to T scores and then used to make up 6 cognitive domains  plus a total score (Processing Speed, Attention and Vigilance, Working Memory, Verbal Learning, Reasoning Speed, Social Cognition, and a Total Score for the Battery).

Thanks again and I appreciate your patience, I'm new to all of this and really enjoy learning from everyone!

V/R

Justin

---

From: Art Kendall [[hidden email]]
Sent: Sunday, September 30, 2012 8:31 AM
To: Justin Blehar
Cc: [hidden email]
Subject: Re: [SPSSX-L] Controlling for Race with SPSS 20

Since you have tiny Ns when grouping cases this way, perhaps you should pool some of the groups? Or use and additional IV? with a 3 way design
As is, it looks like you are double counting cases.
E.g., how do you distinguish heavy smokers from current smokers?

Can you put your cases into a 3 way design (3 * 2 *2), degree of exposure, race as IVs and cognition as the DV?
degree of exposure 3 levels: never, light, heavy
current status 2 levels: current/not
race 2 levels: Caucasian, African American.

How are you measuring cognition?

Art Kendall

Social Research Consultants

On 9/29/2012 8:32 PM, Justin Blehar wrote:

The population is outpatient individuals suffering from schizophrenia. Most of the research I found either does not list race or only has Caucasians in their sample. I have a large enough N to run only Caucasians but this limits how well the results can be generalized. I'm trying to avoid this if possible.

Total Sample

Caucasian - N = 76

African American- N = 44

Groups

Never Smokers: Caucasian N = 15 African American N = 23

Former Smokers: Caucasian N = 29 African American N = 2

Nonsmokers: Caucasian N = 44 African American N = 25

Heavy Smokers: Caucasian N = 9 African American N = 5

Light Smokers: Caucasian N = 23 African American N =14

Current Smokers: Caucasian N = 32 African American N = 19

Thanks for your reply :)

V/R

Justin

________________________________________

From: SPSSX(r) Discussion [[hidden email]] on behalf of Michael Palij [[hidden email]]

Sent: Saturday, September 29, 2012 8:11 PM

To: [hidden email]

Subject: Re: Controlling for Race with SPSS 20

If you control for race in whatever manner, to what population

would your conclusions apply to?  And by population I mean

humans, not a mathematical distribution.

-Mike Palij

New York University

[hidden email]

On Sat, Sep 29, 2012 at 6:18 PM, Justin Blehar [hidden email] wrote:

Hello All,

Not sure how much detail is needed so I'll give you a quick overall but I'm

trying to control for race and am unsure how to best go about this. I know

that I can run a partial correlation and control for race using the menu but

is this really controlling for race? If not is there a better way? How would

I do this for a t-test?

This is a cross sectional design looking at cognition and smoking in a

psychiatric population. There are six groups I'm looking at; Never Smokers,

Former Smokers, Nonsmokers (includes both never smokers and former smokers),

Heavy Smokers, Light Smokers, and Smokers (includes heavy and light

smokers). I have 36 scale variables that I want to compare between each of

these groups. When I break out the groups by race (just looking at box plots

and mean comparisons) there are clearly some large race effects (e.g.

parental education, level of functioning, IQ, etc...). I'd like to be able

to correct for this in each analysis. I'm running both correlations and

t-tests (maybe this isn't the best process?).

If I run a partial correlation and control for race is this really

controlling for race?

When running the t-tests how do I control for race?

Any help would be greatly appreciated.

V/R

Justin

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

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

 

[Steve Simon, P.Mean Consulting]

Justin Blehar wrote:

> Not sure how much detail is needed so I'll give you a quick overall
> but I'm trying to control for race and am unsure how to best go about
> this. I know that I can run a partial correlation and control for
> race using the menu but is this really controlling for race? If not
> is there a better way? How would I do this for a t-test?
>
> This is a cross sectional design looking at cognition and smoking in
> a psychiatric population. There are six groups I'm looking at; Never
> Smokers, Former Smokers, Nonsmokers (includes both never smokers and
> former smokers), Heavy Smokers, Light Smokers, and Smokers (includes
> heavy and light smokers). I have 36 scale variables that I want to
> compare between each of these groups. When I break out the groups by
> race (just looking at box plots and mean comparisons) there are
> clearly some large race effects (e.g. parental education, level of
> functioning, IQ, etc...). I'd like to be able to correct for this in
> each analysis. I'm running both correlations and t-tests (maybe this
> isn't the best process?).

"Controlling for Race" is more properly thought of as risk adjustment.
There are several ways to do this. The simplest is to fit a general
linear model with both race and smoking as independent variables. The
estimates for smoking that you get with the LSMEANS option represents
the estimated average outcome when the mix of race is the same in each
group. To do this well, you should create your own indicator variables
rather than let SPSS do it for you, as SPSS might choose the "wrong"
reference level.

Another way to do this, as you noted, is to restrict your sample to just
one race group, but this is wasteful of the data that you worked so hard
to collect. Still, that might be a nice secondary analysis.

The suggestion to look at interactions is not "controlling for race" but
rather, trying to identify subgroups. It might be, for example, that
smoking shows large differences in the Black subpopulation and not in
the Asian-American subpopulation. I generally discourage looking at
interactions, unless there is a strong a priori belief that
subpopulation effect exist, either based on previous research or based
on some plausible scientific mechanism.

Steve Simon, [hidden email], Standard Disclaimer.
Sign up for the Monthly Mean, the newsletter that
dares to call itself average at www.pmean.com/news

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

[Poes, Matthew Joseph]

Justin I think you need to reconsider this model.  You don’t just have a little issue with power, it’s a huge issue.  If you consider the 6 groups, basically you are saying you want to get an overall effect and specific set of effects for 36 predictors between 6 groups.  Running this through G*Power, I calculate that you would realistically need around 2-300 people.  You might argue that you are going to test these all separate, as seems to be indicated by your request for a t-test, but you really should be then correcting for this as multiple tests (in other words the same rule still applies for power). 

 

I would consider reducing group sizes.  I would also take a look at your 36 variables, and look at them carefully to see if they can be reduced or combined.  I don’t know what these 36 variables are, but you may find that they are indicative of some latent constructs when combined (they may have even been designed that way to begin with).

 

Now, as for how to do what you want, people have given you this already.  You can’t do it with a t-test, you would switch to an ANOVA.  When you “control” for something, you are computing the variance in the Y variable accounted for by the “control” variable.  In reality, you are removing the variance from the beta 1 coefficient for the beta 2 coefficient, which you are calling a control variable.  The reality of this is that the beta 2 coefficient is also now having the variance of beta 1 removed from it, i.e. both are controls for each other. 

 

You can accomplish this in a few ways, but you won’t actually use the ANOVA command under means comparison, you want Univariate general linear model.  You can include the race variable as a fixed effect factor or as a covariate, and in the end, it will give you precisely the same results.  Either place is fine, they are mathematically equivalent.  For the model, you want this to be main effects only.  If you include the interaction, then you end up looking at the specific results within African American vs Caucasian, and not just controlling for the variance explained of being an African American.  That’s a different research question altogether.  So my recommendation would be as follows.

 

UNIANOVA Yvar BY Smokecat WITH RaceX

  /METHOD=SSTYPE(3)

  /INTERCEPT=INCLUDE

  /PRINT=PARAMETER

  /CRITERIA=ALPHA(.05)

  /DESIGN=Smokecat RaceX.

 

The code above includes syntax to give you the parameter estimates.  Makes things easier to interpret, you get the beta’s then.  No interactions, as mentioned.  Your only remaining problem is that Yvar is really Yvar 1 through 36.  Many would argue you need to take your alpha criteria for each, divide by the number of Y’s, and input that.  That would come to .001.  We could get into a long discussion of when this is appropriate and when it isn’t.  If this is part of a true experimental study, and you want the results to be indicative of experimental trials, then you need to do this for anyone to take the results seriously.  If its totally exploratory, then there is an argument that no correction is needed.  I would argue, however, that the results need to be presented in this way, that a follow-up study is necessary with sufficient sample size to allow for this.      

 

Matthew J Poes

Research Data Specialist

Center for Prevention Research and Development

University of Illinois

510 Devonshire Dr.

Champaign, IL 61820

Phone: [phone]

email: [hidden email]

 

 

From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Justin Blehar
Sent: Saturday, September 29, 2012 5:18 PM
To: [hidden email]
Subject: Controlling for Race with SPSS 20

 

Hello All,

Not sure how much detail is needed so I'll give you a quick overall but I'm trying to control for race and am unsure how to best go about this. I know that I can run a partial correlation and control for race using the menu but is this really controlling for race? If not is there a better way? How would I do this for a t-test?

This is a cross sectional design looking at cognition and smoking in a psychiatric population. There are six groups I'm looking at; Never Smokers, Former Smokers, Nonsmokers (includes both never smokers and former smokers), Heavy Smokers, Light Smokers, and Smokers (includes heavy and light smokers). I have 36 scale variables that I want to compare between each of these groups. When I break out the groups by race (just looking at box plots and mean comparisons) there are clearly some large race effects (e.g. parental education, level of functioning, IQ, etc...). I'd like to be able to correct for this in each analysis. I'm running both correlations and t-tests (maybe this isn't the best process?).

If I run a partial correlation and control for race is this really controlling for race?

When running the t-tests how do I control for race?

Any help would be greatly appreciated.

V/R

Justin