Ron...from our buddy Dan...have fun with this one.

________________________________

From: Lofald, Dan [mailto:[EMAIL PROTECTED]
Sent: Mon 7/31/2006 2:00 PM
To: Lehman, David; Ron Carson
Subject: RE: [OTlist] Interesting Stats



Howdy David, Howdy Ron

 

We might not be sitting at the lunch table but the good conversation continues.

As the Irish saying goes, "Is this a private fight or can anyone join"

 

It appears to be that there are about 6 different concepts in play.

 

David's comments about taking all of the variables en' mass and seeing what 
regression would tell us.

Yep-folks do this all of the time.  For example, I was involved trying to find 
lead indicators of newborn deaths at Univ. of Florida Hospital.  There were 
1400 independent variables and 1 dependent variable (death/life).

We hunted around for months looking for best predictive sets of predictors.  
The approach to this work was, "if it works, it works."  However, this is 
nothing more than mainframe level data grubbing.  It can never tell us anything 
scientifically definitive because the alpha level is astronomical.  However, it 
might give insight on a place to look with tools that are more sophisticated.  

 

Invariably, meaningful finds with first-order relationships are rare events.  
The good stuff is going to emerge out of finding important suppressor and 
mediator variables at work.  In addition, most of the variables will probably 
not be in a linear relationship with each other (especially not with human 
beings).  With regression, you can check for suppressor and mediator variables 
and you can test polynomial solutions --- however-to get to the good stuff, we 
would probably have use multi-variate tools.  

 

At the broader level - the two of you have an ontological & mathematical 
conversation combined (I love it).

 

Let's look at the passage, "All humans are unique."  What does that statement 
mean?  Every part of every human is unlike every part of every other human?  If 
that were the case, and we used Set Theory, we would have 121 billion 
non-intersecting circles.  If that were the case, not only would science be 
impossible, but so also preclude the possibility of language and culture.

 

If by, "All humans are unique," we mean, that some part of my circle does not 
overlap with your circle - we are now onto something important.  Now we can ask 
how much of our two Venn circles overlap and do not overlap.  This is 
expression we have with a Pearson product-moment correlation of p = .80, which 
means that 64% of the variance in one variable can be predicted by the variance 
of the other circle.

 

But what happens when there are ten of us.  How do our circles overlap now (we 
are talking about the ephemeral phenomological/psychological variables that you 
guys are talking about).  If we look at enough people can we find patterns 
among what initially appeared discretely idiosyncratic?

 

[[Oh gee -  I cannot finish this response this week but I should would like to 
take it up with you guys another time.  You are drilling down to one of the 
most interesting and important pieces of social science.

 

Dan

 

 

 

Daniel R. Lofald, PhD

Staff Development Coordinator

Chippewa Valley Technical College

620 W. Clairemont Ave., Eau Claire, WI 54701-6162

Phone 715-852-1328 Fax 715-833-6451

 

________________________________

From: Lehman, David [mailto:[EMAIL PROTECTED] 
Sent: Wednesday, July 26, 2006 8:54 PM
To: Lofald, Dan
Subject: FW: [OTlist] Interesting Stats

 

Ron remembers you.....well!

 

More thoughts?

 

________________________________

From: [EMAIL PROTECTED] on behalf of Ron Carson
Sent: Wed 7/26/2006 8:42 PM
To: Lehman, David
Subject: Re: [OTlist] Interesting Stats

Given  that  you  have  a  PhD,  and  if  I  remember,  you and Dan L.
frequently  talked  about  this  stuff,  I  feel  that I am definitely
out-classed!!  <lol>. But, being the brave, (uh, stupid) person that I
am, I will continue wading in the deep end of the pool.

It  seems  that regression analysis looks at individual variables. But
we  know  that people are not just collection of individual variables.
We are after all, greater then the sum of our parts.

How does regression analysis look at the SUM of the variables?

Also,  is it possible to quantify a subjective feeling? I know this is
done  with  pain  scales,  but  the  subjective  nature of such scales
renders  them  almost useless for comparisons sake. In other words, my
reported pain of 9 is total meaningless when COMPARED to your reported
pain of 9.

Lastly, just because someone has a history of any of the variables, in
and  of itself, that history is meaningless. For example, just because
someone  has  a  history  of total hip arthroplasty, that doesn't mean
they will need adaptations to dress their lower body.

Ok, one more "lastly". You said:

david> The  more variables, the more chances for your results occuring
david> because of poor internal validity

Are  you  saying  that  the  greater the number of variables, the less
likely  you  can  predict the results? If so, they I maintain that the
number of variable affecting patient outcomes is infinite and thus you
can  never  truly  predict  the outcomes. Keep in mind that I refer to
patients, not to procedures.

Ron

----- Original Message -----
From: Lehman, David <[EMAIL PROTECTED]>
Sent: Wednesday, July 26, 2006
To:   [email protected] <[email protected]>
Subj: [OTlist] Interesting Stats

LD> I am not surprised you disagree, Ron!  Thats what makes us get along so 
well!

LD> I think you walked right into what I wanted to get across:

LD> Each of the variables you listed below can be measured,
LD> right?  If something (a dependent or attribute variable) can be
LD> measured, then it has a number which can then be used to determine
LD> if the unique variable is significant , and this is the important
LD> part, I think, that when using a regression analysis, it takes
LD> into account all the other unique and not unique variables thought
LD> to be a contributor to the healing (increased functional
LD> activities, increased occupation). 

LD> So, if one individual, has a history of any or most of these,
LD> then how does that effect your treatment versus if she had 30%
LD> relevant, or 10% relevant, etc...  The more variables, the more
LD> chances for your results occuring because of poor internal
LD> validity (poor job controlling all the other variablesvariables) -
LD> and you dont know if it were treatment approach a or b or c or d
LD> or e.....or placebo.
LD> OK...I am going statistical here and trying to explain
LD> it..,.....but, if you can quantify something, you can study its
LD> relationship with other variables and outcomes.

LD> This is fun.

LD> ________________________________

LD> From: [EMAIL PROTECTED] on behalf of Ron Carson
LD> Sent: Wed 7/26/2006 7:31 PM
LD> To: Lehman, David
LD> Subject: Re: [OTlist] Interesting Stats



LD> Hello David:

LD> I  hate to disagree but every person has unique variables. That's what
LD> makes  us  unique. And perhaps the greatest and uniqueness variable of
LD> all  is the subjective feeling and experiences associated with injury,
LD> disease  and  illness. It is in fact that subjective experience of our
LD> patient's that makes being an OT so difficult.

LD> Regarding a list, here's a few:

LD> age,  gender,  marital  status,  employment  history,  race, religion,
LD> medical   history,  current  medications,  spiritual  beliefs,  family
LD> support,   financial   support,   cognitive   status,  mental status,
LD> education,   prior  experience  with  OT,  expectations,  diet,  sleep
LD> patterns, do they have regular bowel movements, etc.

LD> The list is truly endless!!

LD> This is a GREAT topic

LD> Ron

LD> ----- Original Message -----
LD> From: Lehman, David <[EMAIL PROTECTED]>
LD> Sent: Wednesday, July 26, 2006
LD> To:   [email protected] <[email protected]>
LD> Subj: [OTlist] Interesting Stats

LD>> I mis-spoke, perhaps....meaning that there really are no
LD>> unique variables that ONLY exist to one person on earth.  If
LD>> enough data is gathered on a population, the "unique" or should I
LD>> say less appearing variables we call unique, would  be detectable
LD>> and the statistical model would show us what contribution that
LD>> particular "unique variable" makes and if it is significant in the
LD>> outcome.

LD>> I guess my next question would be to ask the members to list
LD>> variables they consider unique to a person that have an impact on
LD>> the treatment they provide.

LD>> ________________________________

LD>> From: [EMAIL PROTECTED] on behalf of Ron Carson
LD>> Sent: Wed 7/26/2006 4:04 PM
LD>> To: Lehman, David
LD>> Subject: Re: [OTlist] Interesting Stats



LD>> I  don't  know  much,  if  anything,  about regression theory, but I'm
LD>> pretty  confident  that  it is impossible to take into account ALL the
LD>> UNIQUE  variables  that  account  for  healing.  Also, it seems rather
LD>> counterintuitive  to  try include a "unique" variable into a "general"
LD>> logarithm.

LD>> Ron

LD>> ----- Original Message -----
LD>> From: Lehman, David <[EMAIL PROTECTED]>
LD>> Sent: Wednesday, July 26, 2006
LD>> To:   [email protected] <[email protected]>
LD>> Subj: [OTlist] Interesting Stats

LD>>> all of the unique variables of
LD>>> a client could one day be analysed in a regression model
LD>>> and thus determine how much these
LD>>> unique variables actually account for healing, then
LD>>> include them in the logaritm?


LD>> --
LD>> Unsubscribe?
LD>>   [EMAIL PROTECTED]

LD>> Change options?
LD>>   www.otnow.com/mailman/options/otlist_otnow.com

LD>> Archive?
LD>>   www.mail-archive.com/[email protected]

LD>> Help?
LD>>   [EMAIL PROTECTED]




LD> --
LD> Unsubscribe?
LD>   [EMAIL PROTECTED]

LD> Change options?
LD>   www.otnow.com/mailman/options/otlist_otnow.com

LD> Archive?
LD>   www.mail-archive.com/[email protected]

LD> Help?
LD>   [EMAIL PROTECTED]




--
Unsubscribe?
  [EMAIL PROTECTED]

Change options?
  www.otnow.com/mailman/options/otlist_otnow.com

Archive?
  www.mail-archive.com/[email protected]

Help?
  [EMAIL PROTECTED]

-- 
Unsubscribe?
  [EMAIL PROTECTED]

Change options?
  www.otnow.com/mailman/options/otlist_otnow.com 

Archive?
  www.mail-archive.com/[email protected]

Help?
  [EMAIL PROTECTED]

Reply via email to