[tips] RE: Thanks! (was: Citation for usefulness of pretests for ANOVA)

[Jim Clark]

Hi

Because an average of percentiles does not correspond to the percentile
of average raw or scaled scores.  Described clearly at:

http://learn.sdstate.edu/erionr/eder711/avgperc.html 

I'm not sure how that issue is relevant to the validity of parametric
analysis of ordinal judgments by individual subjects.

Take care
Jim


James M. Clark
Professor of Psychology
204-786-9757
204-774-4134 Fax
[EMAIL PROTECTED]

>>> [EMAIL PROTECTED] 26-Apr-06 10:32:09 PM >>>
Publishers of standardized achievement tests routinely caution against
averaging percentile ranks.  Instead, they recommend using such
calculations with normalized scaled scores.  Why is that?



Michael T. Scoles, Ph.D.
Associate Professor of Psychology & Counseling
University of Central Arkansas
Conway, AR 72035
>>> [EMAIL PROTECTED] 04/26/06 8:15 PM >>>
Perhaps simple-mindedly, I view Likert type scales of single items as
legitimate candidates for anovas and the like because the responses
are
ordered.  So if one group has a mean of 2.4 and another a mean of 3.6,
we know that the latter group was on average more in agreement with
whatever was rated (depending on orientation of scale, of course). 
This
conclusion is legitimate no matter what the "real" differences between
the 4 levels.



---
To make changes to your subscription go to:
http://acsun.frostburg.edu/cgi-bin/lyris.pl?enter=tips&text_mode=0&lang=english






---
To make changes to your subscription go to:
http://acsun.frostburg.edu/cgi-bin/lyris.pl?enter=tips&text_mode=0&lang=english

[tips] RE: What a brilliant idea!

[Shearon, Tim]

Michael- I think they might see it as an intrusion into their business plan. 
Set it up as a "flea market" kind of thing where students can sell anything 
(well, anything legal, etc!) and they shouldn't be able to complain. Our 
students call their's a bulletin board and use it for announcements and such in 
addition to selling things. Tim

___
Timothy O. Shearon, PhD
Professor and Chair Department of Psychology
Albertson College of Idaho
Caldwell, ID 83605
email: [EMAIL PROTECTED]

teaching: intro to neuropsychology; psychopharmacology; general; history and 
systems




-Original Message-
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]
Sent: Wed 4/26/2006 5:57 PM
To: Teaching in the Psychological Sciences (TIPS)
Subject: [tips] What a brilliant idea!
 
A student at my college has come up with the idea of creating a
website
so that students can list texts for sale.Apparently the company that
buys used texts on campus is being very antagonistic against this
student enterpreuner.Students can log in for free(currently).

Michael J.Sylvester,PhD
Daytona Beach,Florida


---
To make changes to your subscription go to:
http://acsun.frostburg.edu/cgi-bin/lyris.pl?enter=tips&text_mode=0&lang=english


<>---
To make changes to your subscription go to:
http://acsun.frostburg.edu/cgi-bin/lyris.pl?enter=tips&text_mode=0&lang=english

[tips] RE: Thanks! (was: Citation for usefulness of pretests for ANOVA)

[Michael Scoles]

Publishers of standardized achievement tests routinely caution against 
averaging percentile ranks.  Instead, they recommend using such calculations 
with normalized scaled scores.  Why is that?



Michael T. Scoles, Ph.D.
Associate Professor of Psychology & Counseling
University of Central Arkansas
Conway, AR 72035
>>> [EMAIL PROTECTED] 04/26/06 8:15 PM >>>
Perhaps simple-mindedly, I view Likert type scales of single items as
legitimate candidates for anovas and the like because the responses are
ordered.  So if one group has a mean of 2.4 and another a mean of 3.6,
we know that the latter group was on average more in agreement with
whatever was rated (depending on orientation of scale, of course).  This
conclusion is legitimate no matter what the "real" differences between
the 4 levels.



---
To make changes to your subscription go to:
http://acsun.frostburg.edu/cgi-bin/lyris.pl?enter=tips&text_mode=0&lang=english

[tips] Re: Animal/human learning(was Common dense..)

[Michael Scoles]

This is called "word salad."

Michael T. Scoles, Ph.D.
Associate Professor of Psychology & Counseling
University of Central Arkansas
Conway, AR 72035
>>> [EMAIL PROTECTED] 04/28/05 9:15 AM >>>

- Original Message - 
From: "DeVolder Carol L" <[EMAIL PROTECTED]>
To: "Teaching in the Psychological Sciences (TIPS)" 

Sent: Tuesday, April 25, 2006 12:10 PM
Subject: [tips] Re: Common sense revisited

I am glad you clarified this by stating"one does not necessarily seek 
reinforcement" which implies that one can necessarily do.For example 
contestrants on American
idol.Although some principles of animal learning may apply to human 
learning,there are differences.As a matter
of fact underlying cognitive and social learning theories
is the idea of expectancy which means that reinforcement can be made to 
happen.Attempts to equate animal learning to human learning has had 
drawbacks.Way,way back Tolman coined the idea of expectancy in rats-a kind 
of cognitive map- or form semantic networks.The psychology
but the idea could not be supported by experimental research.Then came 
Mowrer with the idea of hope in
animals with was like religious concept.Animals do respond to nonsense 
syllables.At Mizzou I worked under Don Kausler who had a text titled the 
Psychology of Verbal Learning and Memory and this was far beyond
animal conditioning.

Michael SylvesterPhD
Daytona Beach,Florida




---
To make changes to your subscription go to:
http://acsun.frostburg.edu/cgi-bin/lyris.pl?enter=tips&text_mode=0&lang=english



---
To make changes to your subscription go to:
http://acsun.frostburg.edu/cgi-bin/lyris.pl?enter=tips&text_mode=0&lang=english

[tips] Re: Animal/human learning(was Common dense..)

[Michael Sylvester]

- Original Message - 
From: "DeVolder Carol L" <[EMAIL PROTECTED]>
To: "Teaching in the Psychological Sciences (TIPS)" 


Sent: Tuesday, April 25, 2006 12:10 PM
Subject: [tips] Re: Common sense revisited



Without responding to the below-signed poster (not Annette), I'd like to
point out (to anyone in general) that human beings are animals, thus
human behavior IS animal behavior, and so human learning IS animal
learning. If someone continues a behavior, then it is being reinforced.
One doesn't necessarily seek reinforcement--it simply occurs.

I am glad you clarified this by stating"one does not necessarily seek 
reinforcement" which implies that one can necessarily do.For example 
contestrants on American
idol.Although some principles of animal learning may apply to human 
learning,there are differences.As a matter

of fact underlying cognitive and social learning theories
is the idea of expectancy which means that reinforcement can be made to 
happen.Attempts to equate animal learning to human learning has had 
drawbacks.Way,way back Tolman coined the idea of expectancy in rats-a kind 
of cognitive map- or form semantic networks.The psychology
but the idea could not be supported by experimental research.Then came 
Mowrer with the idea of hope in
animals with was like religious concept.Animals do respond to nonsense 
syllables.At Mizzou I worked under Don Kausler who had a text titled the 
Psychology of Verbal Learning and Memory and this was far beyond

animal conditioning.

Michael SylvesterPhD
Daytona Beach,Florida




---
To make changes to your subscription go to:
http://acsun.frostburg.edu/cgi-bin/lyris.pl?enter=tips&text_mode=0&lang=english

[tips] RE: Thanks! (was: Citation for usefulness of pretests for ANOVA)

[Jim Clark]

>>> [EMAIL PROTECTED] 26-Apr-06 11:45:23 AM >>>

MC:
Well, for the first question, the mean is not meaningful because we
don't know about the distance between each of the categories of
response.  At least, that's what I was taught: you cannot take a mean if
you have no measure of distance between your numbers.  Maybe that's
changed; I'm old, but it's always been my understanding that the
requirement for interval or ratio data in most parametric tests is
because we do division and multiplication on the numbers to compute
those stats.  Division and multiplication are not meaningful for ordinal
data.

JC:
I'm old too Marc!  The debate about levels of measurement and
parametric statistics is a long-standing one that surfaces every now and
then, but never appears to be completely settled one way or the other. 
A number of statisticians have argued that Stevens was incorrect in his
classification of measures and that the resulting advice about
parametric statistics was misguided.  Here are a few links with multiple
citations to the debate (I typed "levels of measurement and gaito" into
google (John Gaito was one of Stevens's critics).  Interestingly, I
noticed that a number of recent papers appeared to be arguing once again
for a position like yours.  I didn't read the papers, so don't know to
what extent they have anything new to offer.

http://www.math.yorku.ca/Who/Faculty/Monette/Ed-stat/0098.html 
http://www2.msstate.edu/~jmg1/8803/meas.htm 
http://www.spss.com/research/wilkinson/Publications/Stevens.pdf 

Perhaps simple-mindedly, I view Likert type scales of single items as
legitimate candidates for anovas and the like because the responses are
ordered.  So if one group has a mean of 2.4 and another a mean of 3.6,
we know that the latter group was on average more in agreement with
whatever was rated (depending on orientation of scale, of course).  This
conclusion is legitimate no matter what the "real" differences between
the 4 levels.

With multiple Likert items averaged together (or equivalently summed),
then our measurement will have more gradations (i.e., be more
continuous) and even be more likely to assume a normal distribution
(assuming appropriate variation in the items being summed).  This is
probably a more typical case in psychology.

MC:
The second question about the nature of the distribution has to do with
the requirement that the DV be normally distributed in the population;
and because the DV is ordinal, it cannot be normally distributed.

JC:
This is not correct.  The Central Limit Theorem does not require that
the DV be normally distributed.  The Sampling Distribution of the means
will be normally distributed (or approximately so) even if the samples
are drawn from a non-normal distribution.  The following simple
simulation in SPSS shows that the t-distribution produces the expected
5% of ts > critical value even when samples are drawn from uniform (I
used 1, 2, 3, 4 as in your original example).  If you remark out the
first two computes in the do repeat (the uniform) and unremark the next
two computes (the normal) you will see that approximately 5% of ts are
significant when Ho is true for both normal and uniform distributions,
even with relatively modest sample sizes of 10 in each of two groups. 
If you graph the t (or the difference between means) you will see the
mound-shape that has emerged from averaging.

input program.
loop o = 1 to 1.
do repeat v1 = c1 c2 c3 c4 c5 c6 c7 c8 c9 c10 /v2 = t1 t2 t3 t4 t5 t6
t7 t8 t9 t10.
comp v1 = trunc(rv.uniform(1,5)).
comp v2 = trunc(rv.uniform(1,5)).
*comp v1 = rnd(rv.norm(2.5,1.12)).
*comp v2 = rnd(rv.norm(2.5,1.12)).
end repeat.
end case.
end loop.
end file.
end input program.
comp mc = mean(c1, c2, c3, c4, c5, c6, c7, c8, c9, c10).
comp sdc = sd(c1, c2, c3, c4, c5, c6, c7, c8, c9, c10).
comp mt = mean(t1, t2, t3, t4, t5, t6, t7, t8, t9, t10).
comp sdt = sd(t1, t2, t3, t4, t5, t6, t7, t8, t9, t10).
comp stderr = sqrt( ((sdc**2+sdt**2)/2) * (1/10 + 1/10) ).
comp t = (mc-mt)/stderr.
comp tsig = (t > idf.t(.95,18) ).
freq tsig.

Frequency Percent Valid Percent Cumulative  
Percent 
 Valid .00  9484  94.894.8  94.8
   1.00 516   5.2 5.2   100.0   
   Total1 100.0   100.0 

Take care
Jim


James M. Clark
Professor of Psychology
204-786-9757
204-774-4134 Fax
[EMAIL PROTECTED]



---
To make changes to your subscription go to:
http://acsun.frostburg.edu/cgi-bin/lyris.pl?enter=tips&text_mode=0&lang=english

[tips] What a brilliant idea!

[msylvester]

A student at my college has come up with the idea of creating a
website
so that students can list texts for sale.Apparently the company that
buys used texts on campus is being very antagonistic against this
student enterpreuner.Students can log in for free(currently).

Michael J.Sylvester,PhD
Daytona Beach,Florida


---
To make changes to your subscription go to:
http://acsun.frostburg.edu/cgi-bin/lyris.pl?enter=tips&text_mode=0&lang=english

[tips] Re: BMI,soft drinks and blacks

[Michael Scoles]

I could live my life without knowing some racial differences.

Michael T. Scoles, Ph.D.
Associate Professor of Psychology & Counseling
University of Central Arkansas
Conway, AR 72035
>>> [EMAIL PROTECTED] 04/26/06 6:36 PM >>>

. . . Blacks in general have a faster BM for certain food substances. . .



---
To make changes to your subscription go to:
http://acsun.frostburg.edu/cgi-bin/lyris.pl?enter=tips&text_mode=0&lang=english

[tips] BMI,soft drinks and blacks

[msylvester]

The current thread reminds me of the days I held a seminar in Black
Psychology at an area college.It was assumed then that obesity in
black teens were due to nutritional factors including the consumption
of soft drinks.But BMI varies among races.Blacks in general have a
faster BM for certain food substances.SO assuming that the
experimenters controlled
for racial variables,the conclusions of this study could be viewed as
bias.

Michael Sylvester,PhD
Daytona Beach,Florida


---
To make changes to your subscription go to:
http://acsun.frostburg.edu/cgi-bin/lyris.pl?enter=tips&text_mode=0&lang=english

[tips] Re: BMI and soft drink consumption: data analysis

[Stephen Black]

On 26 Apr 2006 at 14:03, David Epstein offered:

> > 
> Stephen, I have a fabulous little Mac-only application called
> GraphClick
>  
> that can turn a graph back into the numerical data it depicts. 
> I could do that will the scatterplots if it would help you.  

I am moved to announce that within a femtosecond of accepting David's generous 
offer 
(with standard insincere disclaimer, only if it's no trouble, yada yada), the 
reconstituted data 
landed in my computer. That is one killer of an ap!

Sincere (sort-of) disclaimer: I own no stock in this company (maybe I should).

Thanks, David.

Stephen

__
Stephen L. Black, Ph.D.   
Department of Psychology
Bishop's  University  
Lennoxville, QC  J1M 1Z7
Canada
[EMAIL PROTECTED]

Dept web page: www.ubishops.ca/ccc/div/soc/psy
TIPS discussion list for psychology teachers at:
 faculty.frostburg.edu/psyc/southerly/tips/index.htm
__



---
To make changes to your subscription go to:
http://acsun.frostburg.edu/cgi-bin/lyris.pl?enter=tips&text_mode=0&lang=english

[tips] Re: BMI and soft drink consumption: data analysis

[David Epstein]

On Tue, 25 Apr 2006, [EMAIL PROTECTED] went:


The points were all over the place, and there was nothing obviously
different between the groups, at least to my eyeball. The
experimental line did head down with increasing BMI; the control
didn't.

[...]

So I asked them politely for the raw data to do my own analysis.
They refused.


Stephen, I have a fabulous little Mac-only application called
GraphClick
 
that can turn a graph back into the numerical data it depicts. 
I could do that will the scatterplots if it would help you.  Of

course, you still wouldn't have the other covariates.

--David Epstein
  [EMAIL PROTECTED]

---
To make changes to your subscription go to:
http://acsun.frostburg.edu/cgi-bin/lyris.pl?enter=tips&text_mode=0&lang=english

[tips] Re: BMI and soft drink consumption: data analysis

[Rick Froman]

The USA Today headline for Table 2 has got to be: 

STOP DRINKING POP: GET FATTER!

This will go nicely with the recent research that showed a correlation
between drinking diet drinks and gaining weight. It seems that a lot of
overweight people are drinking diet drinks, ergo...

Rick

Dr. Rick Froman
Psychology Department
Box 3055
x7295
[EMAIL PROTECTED] 

Proverbs 14:15 "A simple man believes anything, but a prudent man gives
thought to his steps." 
-Original Message-


Table 2 shows that the intervention group had a mean reduction of 286 
kcal per day for 25 weeks (more than 6 months) and still there was no 
overall significant difference in BMI between the groups! In fact, BMI 
increased in both groups.

Ken

---
Kenneth M. Steele, Ph.D.  [EMAIL PROTECTED]
Department of Psychology  http://www.psych.appstate.edu
Appalachian State University
Boone, NC 28608
USA
---

---
To make changes to your subscription go to:
http://acsun.frostburg.edu/cgi-bin/lyris.pl?enter=tips&text_mode=0&lang=
english


---
To make changes to your subscription go to:
http://acsun.frostburg.edu/cgi-bin/lyris.pl?enter=tips&text_mode=0&lang=english

[tips] Re: BMI and soft drink consumption: data analysis

[Wendi K. Born]

Still, these are "kids" and we can expect that BMI will increase over
any 6-month period of time. During adolescence, body mass changes
dramatically for both boys and girls. In this distribution, two-thirds
of the distribution was below BMI = 25.6. In a distribution of adults,
one now expects two-thirds of the distribution to be above 25. In
addition, people with a BMI of 30 or more typically increase their BMI
over time as well. Slowing the rate of increase in people with
above-normal BMI is considered an important preventive goal, especially
given the difficulty of weight loss and the seeming inevitability of
weight gain. 

Wendi K. Born, Ph.D.
Licensed Clinical Psychologist &
Assistant Professor of Psychology
Baker University
618 8th Street
PO Box 65
Baldwin City, KS 66006-0065
 
(785) 594-8437
[EMAIL PROTECTED]
 

-Original Message-
From: Ken Steele [mailto:[EMAIL PROTECTED] 
Sent: Wednesday, April 26, 2006 11:40 AM
To: Teaching in the Psychological Sciences (TIPS)
Subject: [tips] Re: BMI and soft drink consumption: data analysis

[EMAIL PROTECTED] wrote:
> I'm pondering a recent study by Ebberling (2006) on the effect of 
> reducing soft drink consumption on body mass index (BMI) in
teen-agers. 
> There are some things that bug me about their paper, and I'm looking
for 
> advice and comment from the list, or at least from those of you who
get 
> their jollies from contemplating experimental design and analysis. The

> rest of you will suddenly remember your appointment with the dentist.
> 

Stephen may be correct about the dubious post-hoc nature of the high-BMI

comparisons but I think the most amazing result in the study is in Table

2, if I read it correctly.

Table 2 shows that the intervention group had a mean reduction of 286 
kcal per day for 25 weeks (more than 6 months) and still there was no 
overall significant difference in BMI between the groups! In fact, BMI 
increased in both groups.

I think that this lack of difference show that kids will find other ways

to get calories if sugared soft drinks are eliminated from the diet.


Ken

---
Kenneth M. Steele, Ph.D.  [EMAIL PROTECTED]
Department of Psychology  http://www.psych.appstate.edu
Appalachian State University
Boone, NC 28608
USA
---

---
To make changes to your subscription go to:
http://acsun.frostburg.edu/cgi-bin/lyris.pl?enter=tips&text_mode=0&lang=
english


---
To make changes to your subscription go to:
http://acsun.frostburg.edu/cgi-bin/lyris.pl?enter=tips&text_mode=0&lang=english

[tips] Re: BMI and soft drink consumption: data analysis

[Ken Steele]

[EMAIL PROTECTED] wrote:
I'm pondering a recent study by Ebberling (2006) on the effect of 
reducing soft drink consumption on body mass index (BMI) in teen-agers. 
There are some things that bug me about their paper, and I'm looking for 
advice and comment from the list, or at least from those of you who get 
their jollies from contemplating experimental design and analysis. The 
rest of you will suddenly remember your appointment with the dentist.




Stephen may be correct about the dubious post-hoc nature of the high-BMI 
comparisons but I think the most amazing result in the study is in Table 
2, if I read it correctly.


Table 2 shows that the intervention group had a mean reduction of 286 
kcal per day for 25 weeks (more than 6 months) and still there was no 
overall significant difference in BMI between the groups! In fact, BMI 
increased in both groups.


I think that this lack of difference show that kids will find other ways 
to get calories if sugared soft drinks are eliminated from the diet.



Ken

---
Kenneth M. Steele, Ph.D.  [EMAIL PROTECTED]
Department of Psychology  http://www.psych.appstate.edu
Appalachian State University
Boone, NC 28608
USA
---

---
To make changes to your subscription go to:
http://acsun.frostburg.edu/cgi-bin/lyris.pl?enter=tips&text_mode=0&lang=english

[tips] RE: Thanks! (was: Citation for usefulness of pretests for ANOVA)

[Marc Carter]

Well, for the first question, the mean is not meaningful because we don't know 
about the distance between each of the categories of response.  At least, 
that's what I was taught: you cannot take a mean if you have no measure of 
distance between your numbers.  Maybe that's changed; I'm old, but it's always 
been my understanding that the requirement for interval or ratio data in most 
parametric tests is because we do division and multiplication on the numbers to 
compute those stats.  Division and multiplication are not meaningful for 
ordinal data.

I'm no measurement expert and it's been since grad school since I studied 
meaningfulness, but that seems to me to be important.  

The second question about the nature of the distribution has to do with the 
requirement that the DV be normally distributed in the population; and because 
the DV is ordinal, it cannot be normally distributed.

But again, I welcome correction if I'm teaching wrong stuff.  Let me know!

m

> -Original Message-
> From: Jim Clark [mailto:[EMAIL PROTECTED] 
> Sent: Wednesday, April 26, 2006 11:20 AM
> To: Teaching in the Psychological Sciences (TIPS)
> Subject: [tips] RE: Thanks! (was: Citation for usefulness of 
> pretests for ANOVA)
> 
> Hi
> 
> I would be inclined to analyze this data using either anova 
> or t-test, depending on number of age categories.  A mean of 
> 2.5 means the average was somewhere between "not so much" and 
> "somewhat" and is perfectly interpretable, so I'm not sure 
> what a comment like "we can't really get means from those 
> data" means.  Of course, it helps to use the average of a 
> number of such ratings.  As for the distribution, why could 
> you not determine the distribution from the sample and would 
> it matter anyway?
> 
> Take care
> Jim
> 
> James M. Clark
> Professor of Psychology
> 204-786-9757
> 204-774-4134 Fax
> [EMAIL PROTECTED]
> 
> >>> [EMAIL PROTECTED] 26-Apr-06 9:27:55 AM >>>
> Hey, All --
> 
> Apropos of this, I just this morning showed my methods class 
> an analysis of a Likert-type D.V. ("A lot," "somewhat," "not 
> so much," and "not at all") by an age category I.V.  I asked, 
> "Why do we not take a mean of the ratings and run ANOVA?"
> 
> They said (well, not *all* of them said, but *some* of them 
> said) "Because that's not a good DV for ANOVA."  I asked why, 
> and got two responses (after only a little prompting): "we 
> can't really get means from those data," and "we don't know 
> how it's distributed."  I was pleased.  I then regaled them 
> with the wonders of the Chi-square test of independence.
> 
> Sometimes they're learning more than I think they're learning.  
> 
> (Of course, next month marks the end of a *year* that I've 
> been hammering this stuff into them.)
> 
> Cheers,
> 
> m
> 
> > -Original Message-
> > From: Scheuchenpflug
> > [mailto:[EMAIL PROTECTED]
> > Sent: Wednesday, April 26, 2006 5:54 AM
> > To: Teaching in the Psychological Sciences (TIPS)
> > Subject: [tips] Thanks! (was: Citation for usefulness of 
> pretests for 
> > ANOVA)
> > 
> > Dear Tipsters,
> > 
> > as always, I am amazed by the speed and accuracy of 
> responses on this 
> > list; thank you so much, Stephen and Mike!
> >  
> > I hope the quotation will drive home a point with a 
> subpopulation of 
> > our students who continue to worry about assumptions of ANOVA while 
> > simultaneously endorsing ad-hoc 4-point rating scales as dependent 
> > variables .
> > 
> > Regards,
> > Rainer
> > 
> > Dr. Rainer Scheuchenpflug
> > Lehrstuhl für Psychologie III
> > Röntgenring 11
> > 97070 Würzburg
> > Tel:   0931-312185
> > Fax:   0931-312616
> > Mail:  [EMAIL PROTECTED]
> > 
> > 
> 
> ---
> To make changes to your subscription go to:
> http://acsun.frostburg.edu/cgi-bin/lyris.pl?enter=tips&text_mo
> de=0&lang=english 
> 
> 
> 
> 
> 
> 
> ---
> To make changes to your subscription go to:
> http://acsun.frostburg.edu/cgi-bin/lyris.pl?enter=tips&text_mo
> de=0&lang=english
> 
> 

---
To make changes to your subscription go to:
http://acsun.frostburg.edu/cgi-bin/lyris.pl?enter=tips&text_mode=0&lang=english

[tips] RE: Data sharing re BMI and soft drink consumption:

[Wendi K. Born]

I am not sure if this will help or not, but when I worked in a medical
center, I learned that medical researchers always want to divide their data
into categories, rather than use it as continuous data. I often wondered about their
reasoning. Typically, they said they either thought of the categories as
important clinical groups (a bad argument, usually) or the distribution of data
was unusual and would have required a transformation (better). However, any way
to find “significance” was favored most of the time. 

 

With obesity, medical researchers often use the following rough categorizations
for ADULTS:

BMI 18-25 = normal weight

BMI 26-29 = clinically overweight

BMI 30+ = obese

 

Sometimes, depending on the population, they further break down the obese
category into

BMI 35-39 = very obese

BMI 40+ = morbidly obese

 

I do not think there is a very defensible reason to use tertiles, as
you say, especially with adolescents. I glanced through the article. Re:

"Among the subjects in the upper baseline-BMI tertile (>
25.6)...BMI change differed markedly between the intervention...and the
control...(p= .03), whereas no significant group difference was seen for the
subjects in the middle and lower tertiles (p =.04 for interaction). 

 

I think they may have discussed their interaction in relation to the
tertiles because people who read medical journals do not have training in
thinking about interactions. I think this was their way of explaining that,
within the intervention group, weight-loss increased as a positive function of
baseline BMI. They used the categories and the difference tests as a way of
discussing simple effects. I am just guessing based on my experience doing
research in a preventive medicine department. The way they think of statistics
is less complex, more “creative,” and takes some getting used to.
They live and die by grants and pubs, so there is a huge incentive to find “:significance,”
which is not even well understood.

 

Stephen

 

 

Wendi K. Born, Ph.D.

Licensed Clinical Psychologist &

Assistant Professor of Psychology

Baker
 University

618 8th Street

PO Box
 65

Baldwin City,
 KS 66006-0065

 

(785) 594-8437

[EMAIL PROTECTED]

 




---
To make changes to your subscription go to:
http://acsun.frostburg.edu/cgi-bin/lyris.pl?enter=tips&text_mode=0&lang=english

[tips] RE: Thanks! (was: Citation for usefulness of pretests for ANOVA)

[Jim Clark]

Hi

I would be inclined to analyze this data using either anova or t-test, 
depending on number of age categories.  A mean of 2.5 means the average was 
somewhere between "not so much" and "somewhat" and is perfectly interpretable, 
so I'm not sure what a comment like "we can't really get means from those data" 
means.  Of course, it helps to use the average of a number of such ratings.  As 
for the distribution, why could you not determine the distribution from the 
sample and would it matter anyway?

Take care
Jim

James M. Clark
Professor of Psychology
204-786-9757
204-774-4134 Fax
[EMAIL PROTECTED]

>>> [EMAIL PROTECTED] 26-Apr-06 9:27:55 AM >>>
Hey, All --

Apropos of this, I just this morning showed my methods class an analysis of a 
Likert-type D.V. ("A lot," "somewhat," "not so much," and "not at all") by an 
age category I.V.  I asked, "Why do we not take a mean of the ratings and run 
ANOVA?"

They said (well, not *all* of them said, but *some* of them said) "Because 
that's not a good DV for ANOVA."  I asked why, and got two responses (after 
only a little prompting): "we can't really get means from those data," and "we 
don't know how it's distributed."  I was pleased.  I then regaled them with the 
wonders of the Chi-square test of independence.

Sometimes they're learning more than I think they're learning.  

(Of course, next month marks the end of a *year* that I've been hammering this 
stuff into them.)

Cheers,

m

> -Original Message-
> From: Scheuchenpflug 
> [mailto:[EMAIL PROTECTED] 
> Sent: Wednesday, April 26, 2006 5:54 AM
> To: Teaching in the Psychological Sciences (TIPS)
> Subject: [tips] Thanks! (was: Citation for usefulness of 
> pretests for ANOVA)
> 
> Dear Tipsters,
> 
> as always, I am amazed by the speed and accuracy of responses 
> on this list; thank you so much, Stephen and Mike!
>  
> I hope the quotation will drive home a point with a 
> subpopulation of our students who continue to worry about 
> assumptions of ANOVA while simultaneously endorsing ad-hoc 
> 4-point rating scales as dependent variables . 
> 
> Regards,
> Rainer
> 
> Dr. Rainer Scheuchenpflug
> Lehrstuhl für Psychologie III
> Röntgenring 11
> 97070 Würzburg
> Tel:   0931-312185
> Fax:   0931-312616
> Mail:  [EMAIL PROTECTED] 
> 
> 

---
To make changes to your subscription go to:
http://acsun.frostburg.edu/cgi-bin/lyris.pl?enter=tips&text_mode=0&lang=english 






---
To make changes to your subscription go to:
http://acsun.frostburg.edu/cgi-bin/lyris.pl?enter=tips&text_mode=0&lang=english

[tips] Re: 7.80/5.00 BMI and soft drink consumption: data analysis

[Jim Clark]

Hi

I've had a quick look at the paper, and feel they are on reasonable
grounds, although they present the results in a way that would probably
differ somewhat from the approach in psychology.

A standard multiple regression approach to a study like this would be
to regress post-BMI (or BMI-change, as in their study) on three
predictors: pre-BMI, a categorical predictor for the intervention, and
an interaction term (essentially the product of the first two predictors
or, often preferred, centered [mean subtracted] versions of those
predictors).  A significant effect for the interaction would indicate
that the slopes of the separate regression lines for the two groups
differ significantly.  This is perhaps what their p = .04 for the
interaction refers to.  The significance of the categorical predictor
would indicate whether the intercepts of the two lines differed, which
may or may not be interpretable depending on whether they centered the
continuous pre-BMI measure.

SB:
They analyzed their results using multiple linear regression, with
which 
I'm not familiar. Nevertheless the result was clear: a small, non-
significant decrease in BMI for the experimental group.

JC:
That the mean BMI-change does not differ significantly from 0 is NOT
inconsistent with a significant interaction, or even with a significant
difference between the two groups (it was not I believe).  In fact it
appears that overall both groups increased BMI but the treatment group
increased less.  Nonetheless, there was an interaction indicating
BMI-change varied as a function of the combined influence of pre-BMI and
the intervention.  Such an interaction calls for follow-up analyses.

The regression equivalent of simple effects can be done as the simple
effects of pre-BMI on change for the two groups, which were reported in
the foot note of their graph.  A highly significant decrease in BMI
change with pre-BMI for the intervention group, and no effect for the
control group.  Change goes from positive to negative with increasing
BMI for the intervention group and stays flat (it actually increases
slightly) for the control group.  As Stephen noted, this addresses the
issue of regression to mean.

Simple effects, however, can also be done in the other direction; that
is, simple effects of group at different levels of the pre-BMI measure. 
But this is trickier because of the continuous nature of pre-BMI.  So a
standard approach is to divide up the continuous measure, or plot the
confidence intervals as in their graph.  The CI dips below 0 at a
certain BMi level, indicating a significant difference between control
and treatment emerging at that point.

Even if this second analysis is dubious in some people's minds, the
separate regressions for control and intervention make very clear that
the intervention has markedly changed the relationship between pre-BMI
and change in BMI (or equivalently post-BMI).

Take care
Jim


James M. Clark
Professor of Psychology
204-786-9757
204-774-4134 Fax
[EMAIL PROTECTED]



---
To make changes to your subscription go to:
http://acsun.frostburg.edu/cgi-bin/lyris.pl?enter=tips&text_mode=0&lang=english

[tips] Re: "Cure rate" for mental disorders

[Paul Brandon]

At 9:00 AM -0400 4/26/06, Pollak, Edward  wrote:

Beth's student wrote
"  have a lot of people with mental disorders suffered 
from a head injury before the  onset of symptoms?"


My answer to this part of the question would be "Probably not. 
However, there is an mounting body of (admittedly still 
circumstantial) evidence that many patients may have suffered subtle 
brain damage from viruses (e.g., schizophrenia) or bacteria (e.g., 
streptococcus infection in OCD) and these effects, (interacting with 
genetic predisposition & situational factors) might turn out to be 
far more ubiquitous causes of behavior disorders than most people 
imagine. "


The problem is that this is the sort of background condition that you 
can find in almost anyone if you look for it.

--
The best argument against Intelligent Design is that fact that
people believe in it.

* PAUL K. BRANDON[EMAIL PROTECTED]  *
* Psychology Dept   Minnesota State University  *
* 23 Armstrong Hall, Mankato, MN 56001 ph 507-389-6217  *
*http://krypton.mnsu.edu/~pkbrando/ *

---
To make changes to your subscription go to:
http://acsun.frostburg.edu/cgi-bin/lyris.pl?enter=tips&text_mode=0&lang=english

[tips] RE: Data sharing re BMI and soft drink consumption:

[sblack]

Thanks for the early returns on my query regarding the Ebberling study. I 
have contacted _Pediatrics_ to ask them whether they can help in prying 
loose the data, and I'm waiting for their reply. However, Miguel Roig's 
comments, both on NIH policy and on refusal to share as a general problem 
are both very helpful. As it happens, this is NIH-supported research, and 
I will be pointing this out to them. 

I don't think that regression to the mean is a valid criticism here, as a 
few have mentioned to me. This is because the results are the changes in 
BMI compared with a control group, so regression to the mean should be 
similar in both. What I am sure about is that arbitrarily dividing their 
sample into thirds is not a defensible procedure.

Where I'm having a problem is in their other assertion (the one omitted 
in the abstract) about  the intervention being significant above a BMI of 
30 based on a regression line on differences between the study groups. .  
When I come across something like this, I first assume they know what 
they're talking about, only I'm too dense to follow. But later I start to 
wonder whether the problem is really mine or theirs. The difficulty is 
that it's hard to criticize what they've done if you can't understand it. 
And sometimes I wonder whether this is what they're counting on.  In such 
a case, my preference would be to ignore what I can't follow, and focus 
on what is clearly wrong. But I'm sure in their reply they would just 
then cite the other finding, if that is what it is. 

Keep those helpful comments coming! But please be cautious. Remember, all 
of this is all-too-public.

Stephen

-
Stephen L. Black, Ph.D.  
Department of Psychology 
Bishop's Universitye-mail:  [EMAIL PROTECTED]
Lennoxville, QC J1M 1Z7
Canada

Dept web page at http://www.ubishops.ca/ccc/div/soc/psy
TIPS discussion list for psychology teachers at
http://faculty.frostburg.edu/psyc/southerly/tips/index.htm
---

---
To make changes to your subscription go to:
http://acsun.frostburg.edu/cgi-bin/lyris.pl?enter=tips&text_mode=0&lang=english

[tips] RE: Thanks! (was: Citation for usefulness of pretests for ANOVA)

[Marc Carter]

Hey, All --

Apropos of this, I just this morning showed my methods class an analysis of a 
Likert-type D.V. ("A lot," "somewhat," "not so much," and "not at all") by an 
age category I.V.  I asked, "Why do we not take a mean of the ratings and run 
ANOVA?"

They said (well, not *all* of them said, but *some* of them said) "Because 
that's not a good DV for ANOVA."  I asked why, and got two responses (after 
only a little prompting): "we can't really get means from those data," and "we 
don't know how it's distributed."  I was pleased.  I then regaled them with the 
wonders of the Chi-square test of independence.

Sometimes they're learning more than I think they're learning.  

(Of course, next month marks the end of a *year* that I've been hammering this 
stuff into them.)

Cheers,

m

> -Original Message-
> From: Scheuchenpflug 
> [mailto:[EMAIL PROTECTED] 
> Sent: Wednesday, April 26, 2006 5:54 AM
> To: Teaching in the Psychological Sciences (TIPS)
> Subject: [tips] Thanks! (was: Citation for usefulness of 
> pretests for ANOVA)
> 
> Dear Tipsters,
> 
> as always, I am amazed by the speed and accuracy of responses 
> on this list; thank you so much, Stephen and Mike!
>  
> I hope the quotation will drive home a point with a 
> subpopulation of our students who continue to worry about 
> assumptions of ANOVA while simultaneously endorsing ad-hoc 
> 4-point rating scales as dependent variables . 
> 
> Regards,
> Rainer
> 
> Dr. Rainer Scheuchenpflug
> Lehrstuhl für Psychologie III
> Röntgenring 11
> 97070 Würzburg
> Tel:   0931-312185
> Fax:   0931-312616
> Mail:  [EMAIL PROTECTED]
> 
> 

---
To make changes to your subscription go to:
http://acsun.frostburg.edu/cgi-bin/lyris.pl?enter=tips&text_mode=0&lang=english

[tips] RE: Data sharing re BMI and soft drink consumption:

[Mike Palij]

On Wed, 26 Apr 2006 05:16:44 -0700, Miguel Roig wrote:
>Not really. Data sharing is encouraged in most of the biomedical sciences.
>However, a significant number of  scientists routinely deny requests for
>sharing. It seems to me that if the study was supported by a federal grant,
>the authors have an even greater obligation to share their data. See NIH's
>statement
http://grants.nih.gov/grants/guide/notice-files/NOT-OD-03-032.html.
>For example, the NSF http://www.nsf.gov/pubs/2001/gc101/gc101rev1.pdf
states:
>
>Sharing of Findings, Data, and Other Research Products
>a. NSF expects significant findings from research and education
>activities it supports to be promptly submitted for publication, with
>authorship that accurately reflects the contributions of those
>involved. It expects investigators to share with other researchers,
>at no more than incremental cost and within a reasonable time,
>the data, samples, physical collections and other supporting materials
>created or gathered in the course of the work. It also encourages
>awardees to share software and inventions or otherwise act to
>make the innovations they embody widely useful and usable.
>b. Adjustments and, where essential, exceptions may be allowed to
>safeguard the rights of individuals and subjects, the validity of
>results, or the integrity of collections or to accommodate legitimate
>interests of investigators.
>
>A recent survey by Vogeli, et al. (2006) shows how widespread
>the problem is.
>
>Vogeli, C.D, Yucel, R., Bendavid, E., Jones, L. M., Anderson, M. S., Louis,
>K. S., Campbell, E.G., (2006). Withholding and the Next Generation of
>Scientists: Results of a National Survey. Academic Medicine, 81, 128-136.

>The study can be accessed for free via this rather long URL:
http://www.academicmedicine.org/pt/re/acmed/pdfhandler.1888-200602000-00
005.pdf;jsessionid=EPf1xX7Au1OlcZDOsVVvcRv2psozCb2ZDDPkCZJD1DT3BnL5Y9gU!3895
95325!-949856144!9001!-1

Or one can use the following tiny URL:
http://tinyurl.com/el8hm

My 2 cents US:  even if the APA or the feds request or requires
that researchers make their original data available to others, if has
been my experience that, for whatever reason, many researchers are
incredibly resistant to sharing even if there had been federal funding
for the research.  I think that there are a variety of reasons for this
resistance ranging from the mundane (problems in having the complete
data available in form that can be given to another person in a simple
and convenient form, e.g., in an SPSS or SAS text based program + inline
data file) to more profound (e.g., fear that the requester will write
papers based on the the dataset that will either "scoop" the original
researcher or address issues that the original researcher would not
address [or perhaps wasn't even aware of]).  In one situation in which
I was involved, a federal grant was funding the operation of a research
center which had one of its goals the development of repository for the
data collected by its participant researhers (however, these
participants were funded by their own seperate federal grants) --
no one was really interested in sharing their data and not much
came of this effort.

-Mike Palij
New York University
[EMAIL PROTECTED]

>Miguel







---
To make changes to your subscription go to:
http://acsun.frostburg.edu/cgi-bin/lyris.pl?enter=tips&text_mode=0&lang=english

[tips] "Cure rate" for mental disorders

[Pollak, Edward]

Beth's student wrote
"  have a lot of people with mental disorders suffered from a head 
injury before the  onset of symptoms?"
 
My answer to this part of the question would be "Probably not. However, there 
is an mounting body of (admittedly still circumstantial) evidence that many 
patients may have suffered subtle brain damage from viruses (e.g., 
schizophrenia) or bacteria (e.g., streptococcus infection in OCD) and these 
effects, (interacting with genetic predisposition & situational factors) might 
turn out to be far more ubiquitous causes of behavior disorders than most 
people imagine. "
 
Ed
 
 
 
Edward I. Pollak, Ph.D. 
Department of Psychology, 
West Chester University of Pennsylvania 
Web site: http://mywebpages.comcast.net/epollak/home.htm 
  
Husband, father, grandfather, biopsychologist, bluegrass fiddler, and 
herpetoculturist. in approximate order of importance.

---
To make changes to your subscription go to:
http://acsun.frostburg.edu/cgi-bin/lyris.pl?enter=tips&text_mode=0&lang=english

[tips] RE: Data sharing re BMI and soft drink consumption:

[Miguel Roig]

> So I asked them politely for the raw data to do my own analysis. They
> refused. Given that the APA explicitly tells its authors to fill such
> requests promptly and cooperatively (or something like that), that
> doesn't seem too welcoming of them. does it?

Yes, but they didn't publish in an APA journal; I suppose other
"sciences" have different criteria.

---

Not really. Data sharing is encouraged in most of the biomedical sciences.
However, a significant number of  scientists routinely deny requests for
sharing. It seems to me that if the study was supported by a federal grant,
the authors have an even greater obligation to share their data. See NIH's
statement
http://grants.nih.gov/grants/guide/notice-files/NOT-OD-03-032.html. For
example, the NSF http://www.nsf.gov/pubs/2001/gc101/gc101rev1.pdf states:

Sharing of Findings, Data, and Other Research Products
a. NSF expects significant findings from research and education activities
it supports to be
promptly submitted for publication, with authorship that accurately reflects
the contributions of
those involved. It expects investigators to share with other researchers, at
no more than incremental cost and within a reasonable time, the data,
samples, physical collections and other supporting materials created or
gathered in the course of the work. It also encourages awardees to share
software and inventions or otherwise act to make the innovations they embody
widely useful and usable.
b. Adjustments and, where essential, exceptions may be allowed to safeguard
the rights of
individuals and subjects, the validity of results, or the integrity of
collections or to accommodate legitimate interests of investigators.

A recent survey by Vogeli, et al. (2006) shows how widespread the problem
is.

Vogeli, C.D, Yucel, R., Bendavid, E., Jones, L. M., Anderson, M. S., Louis,
K. S., Campbell, E.G., (2006). Withholding and the Next Generation of
Scientists: Results of a National Survey. Academic Medicine, 81, 128-136.

The study can be accessed for free via this rather long URL:
http://www.academicmedicine.org/pt/re/acmed/pdfhandler.1888-200602000-00
005.pdf;jsessionid=EPf1xX7Au1OlcZDOsVVvcRv2psozCb2ZDDPkCZJD1DT3BnL5Y9gU!3895
95325!-949856144!9001!-1


Miguel




---
To make changes to your subscription go to:
http://acsun.frostburg.edu/cgi-bin/lyris.pl?enter=tips&text_mode=0&lang=english

[tips] Thanks! (was: Citation for usefulness of pretests for ANOVA)

[Scheuchenpflug]

Dear Tipsters,

as always, I am amazed by the speed and accuracy of responses on this list; 
thank you so much, Stephen and Mike!
 
I hope the quotation will drive home a point with a subpopulation of our
students who 
continue to worry about assumptions of ANOVA while simultaneously endorsing
ad-hoc 4-point rating scales as dependent variables . 

Regards,
Rainer

Dr. Rainer Scheuchenpflug
Lehrstuhl für Psychologie III
Röntgenring 11
97070 Würzburg
Tel:   0931-312185
Fax:   0931-312616
Mail:  [EMAIL PROTECTED]




 
Subject: Re: Citation for usefulness of pretests for ANOVA
From: "Stephen Black" <[EMAIL PROTECTED]>
Date: Tue, 25 Apr 2006 10:17:54 -0400


I love it when I can answer questions like this. It was George Box who
provided that delightful simile. 
They just don't write about statistics like that any more, do they?  The
reference is:

Box, G.E.P. (1953). Non-normality and tests on variances. Biometrika 40:
318-35. 

I searched my files but couldn't come up with a copy of the paper. But I
found something on- line nearly as 
good: a "Citation Classic" reminiscence by Box about the circumstances in
which this paper came to be. (A 
"Citation Classic" is an essay which the authors of highly- cited works are
asked to provide].

Box gives the famous quotation there as "To make the preliminary test on
variances is rather like putting to sea in a rowing boat to find out whether
conditions are sufficiently calm for an ocean liner to leave port". That's
remarkably close to what Rainer remembered. 
Pretty good!

The Citation Classic of Box is in _Current Contents_, no. 4 January 25, 1982
and is available on-line at: 
http://www.garfield.library.upenn.edu/classics1982/A1982MX2941.pdf

Note to editors: the essay concludes with a wonderful anecdote. Box says the
referee at Biometrika harshly rejected his paper. But the envelope included
a hand-written note from E.S. Pearson, which said "You will see that the
referee does not like your paper but I do and subject to mild revision I am
going to publish it anyway". Box concludes by noting that papers with novel
ideas are the hardest to publish. 

Stephen

__
Stephen L. Black, Ph.D.   
Department of Psychology
Bishop's  University  
Lennoxville, QC  J1M 1Z7
Canada
[EMAIL PROTECTED]

Dept web page: www.ubishops.ca/ccc/div/soc/psy TIPS discussion list for
psychology teachers at:
 faculty.frostburg.edu/psyc/southerly/tips/index.htm
__

Subject: Re: Citation for usefulness of pretests for ANOVA
From: "Mike Palij" <[EMAIL PROTECTED]>
Date: Tue, 25 Apr 2006 11:12:28 -0400


For those of you with Jstor.org access, the article is avaiable at:
http://links.jstor.org/sici?sici=0006-3444%28195312%2940%3A3%2F4%3C318%3ANAT
OV%3E2.0.CO%3B2-J
or
http://tinyurl.com/zetto


-Mike Palij
New York University
[EMAIL PROTECTED]






---
To make changes to your subscription go to:
http://acsun.frostburg.edu/cgi-bin/lyris.pl?enter=tips&text_mode=0&lang=english