Re: dealing with confounders and residuals

Posted by David Marso on Oct 09, 2011; 9:33am
URL: http://spssx-discussion.165.s1.nabble.com/dealing-with-confounders-and-residuals-tp4881181p4884735.html

If your "various menstrual cycle response variables" are correlated then there are numerous advantages to using Multivariate Multiple Regressions over a series of Univariate Multiple Regressions (Power, Type II Error, understanding the dimensionality of the response hyperspace...).  My referring to MANOVA was perhaps a misnomer (I think of MANOVA really as an umbrella for the GLM of which regression, anova, manova, mancova etc are special cases).  SEM is sort of the granddaddy of all of this .

In SPSS you would specify:
GML dependent variable list WITH independent list (See FM for other specs including saving residuals).
You will see the B coefficients are identical for the Univariate regressions and the Multivariate regressions however the Univariate tests and Multivariate tests will typically vary if the dependent variables are correlated.
Here is a nice presentation of Multivariate Regression.
http://www.google.com/url?sa=t&source=web&cd=14&ved=0CDUQFjADOAo&url=http%3A%2F%2Fwww.psych.yorku.ca%2Flab%2Fpsy6140%2Flectures%2FMultivariateRegression2x2.pdf

Ian Martin-2 wrote

Thanks, I think....

Yes, there are some well supported hypotheses about the role of organic contaminants and heavy metals as endocrine disruptors.

Perhaps I'm not understanding your suggestion, but note that I have only one group, not a control/exposure design?

regards,
Ian

On Oct 7, 2011, at 9:36 PM, David Marso wrote:

> I suggest (as a start) you look into MANCOVA (you really have a multivariate
> problem here and I'll bet the MCV's are correlated.  Two stage analysis
> using PC's as predictors was archaic 20+ years ago when I was studying SEM
> models.  Do you have theoretical hypotheses supported by previous research
> or is this a fishing expedition?  You could do much more efficient and
> directed analyses if you were to apply the appropriate tools.
>
> Ian Martin-2 wrote:
>>
>> I'm looking at the possible association between environmental contaminants
>> (heavy metals and PCBs in human blood) and various menstrual cycle
>> response variables in the study participants.  Obscuring these possible
>> relationships are some confounding variables shown to influence the
>> menstrual cycle dependent variables: age, caffeine consumption, alcohol
>> consumption, etc.
>>
>> I have  run a regression analysis with one of the menstrual cycle
>> variables (dependent) and the suite of confounders, and then saved the
>> residuals.  In a regression on the residuals  (dependent) with the blood
>> contaminant variables, I get some interesting and significant regression
>> coefficients and an overall significant regression model.   I should also
>> mention that I've used PCA on the suite of contaminant concentrations to
>> create 6 uncorrelated predictor variables.
>>
>> Is this a reasonable approach, or is there a better way to block out the
>> effects of the confounders to see the effects of contaminants on the
>> dependent variables?
>>
>> best regards,
>> Ian
>> ____________________
>> Ian D. Martin, Ph.D.
>>
>> Data Analysis &
>> Environmental Consulting
>> _________________________________________________________________
>> This e-mail and any attachments may contain confidential information.  If
>> you are not the intended recipient, please notify the sender  immediately
>> by return e-mail, delete this e-mail and destroy any  copies. Any
>> dissemination or use of this information by a person  other than the
>> intended recipient is unauthorized and may be illegal.
>>
>> Ce message électronique et les fichiers qui y sont joints peuvent
>> contenir des renseignements confidentiels. Si vous n'êtes pas le
>> destinataire visé, veuillez en aviser immédiatement l'expéditeur en
>> répondant à  ce message; effacez ensuite le message et détruisez toute
>> copie. La diffusion ou l'usage de ces renseignements par une personne
>> autre que le destinataire visé n'est pas autorisé et peut constituer  un
>> acte illégal.
>> _________________________________________________________________
>>
>> =====================
>> 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
>>
>
>
> --
> View this message in context: http://spssx-discussion.1045642.n5.nabble.com/dealing-with-confounders-and-residuals-tp4881181p4881812.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