Newsgroup sci.stat.consult 21710

Directory

it won't be long before the entire business of publishing scholarly works
... will be drastically altered given the web and internet. the notion of
reviewed submissions will be outmoded (for good or for bad) soon .... and
papers etc. will merely be posted to the web ... for all to
use/read/take/absorb ...
the day of the printed/referred journal is rapidly limited ...
At 04:57 PM 12/18/96 -0500, you wrote:
>Dear list members:
>
>Ron LaPorte at Univ. of Pittsburgh posted this on the epidemio-l list server
>at Univ. of Montreal and asked that it be forwarded to other lists.  The
>posting describes a Web home page that is addressing the challenges journal
>houses face with regard to publishing on the internet.  In view of Ron's
>request, I have forwarded it to the following list servers:
>
>icrher@listserv.bcm.tmc.edu
>DOSE-NET@orau.gov
>MEDPHYS@CMS.CC.WAYNE.EDU
>cdn-nucl-l@listserv.cis.mcmaster.ca
>radsafe@romulus.ehs.uiuc.edu
>EPIWORLD@UNIVSCVM.CSD.SCAROLINA.EDU
>stat-l@vm1.mcgill.ca
>
>Please don't send it on to these lists.
>
>Leif E. Peterson, Ph.D.
>ICRHER List Administrator (icrher@listserv.bcm.tmc.edu)
>International Consortium for Research on Health Effects of Radiation
>Baylor College of Medicine
>Houston, Texas
>peterson@bcm.tmc.edu
>
>
>Message follows:
>-----------------------------------------
>Date: Tue, 17 Dec 1996 18:31:20 -0400 (EDT)
>From: "Ronald E. LaPorte from Pittsburgh" 
>To: epidemio-l@CC.UMontreal.CA
>Subject: Re: EPIDEMIO-L digest 804
>Message-ID: <01ID4K8JA5TU936DOO@vms.cis.pitt.edu>
>
>
>Dec. 1996
>
>
>
>High Noon for Biomedical Journals
>
>Scientists  from the Global Health Network
>(www.pitt.edu/HOME/GHNet/GHNet.html) predict that within
>5 years most scientists will move their intellectual properties
>to the Internet.  This will spell the demise of most paper
>journals as we know them.  In two recent communications in
>the British Medical Journal they indicated that an Internet
>based system would be much more powerful and available to
>scientists then journals.  Moreover, they are questioning the
>current copyright practice of the journals as this inhibits the
>use of the Internet for posting communications.  A major
>problem, however, is that little is known about how best to
>present research communications on the Internet.  In their
>web site (www.pitt.edu/HOME/GHNet/publications/assassin/),
>an experiment is being conducted where a research
>communication called Scientists Assassinate Journals is
>presented in English, Spanish, Portuguese and Japanese.
>This is presented in a lay version, scientific version, or an
>editor version.  In addition, it is presented in a "hypertext
>comic book form", all include sound.  Within each version
>there are considerable opportunities to provide constructive
>comments concerning the presentation or content.
>
>We encourage scientists, editors, and lay people from all
>walks of life to come to our site and comment.  In this
>manner, we will have data from the scientific community as to
>how best to present scientific research communications.  We
>would suggest  that you forward this to your friends and to list
>servers and news groups as this affects the total scientific
>community, therefore, the more input the better.
>
>Ronald LaPorte, Ph.D.
>Deborah Aaron, Ph.D.
>Akira Sekikawa, M.D.
>Ingrid Libman, M.D., Ph.D.
>Benjamin Acosta, M.D.
>Lucia Iochida, Ph.D.
>Eugene Boostrom, M.D.
>Anthony Villasenor, B.S.
>Amy Brenen, B.S.
>
>

Hello,
If the entries of a matrix W follow the inverse Wishart distribution
with scale matrix S and degrees of freedom v, what is the marginal
distribution of a diagonal element w(i,i) ? what is the distribution of
w(i,i) conditinal on the rest of the entries in W?
Thank you in advance for any hints!
Andreas Hofer

+-------------------+-------------------------------------------------------+
|                   | ******                         *******            **  |
| ***  European     | * ** *                          **   *                |
| *    Course       |   **   ** ****  *****  ******   ** *   ******    ***  |
| **   In           |   **    *** ** **   ** **   **  ****   **   **    **  |
| *    Tropical     |   **    **     **   ** **   **  ** *   **   **    **  |
| ***  Epidemiology |   **    **     **   ** ******   **   * ******     **  |
|                   |  ****   **      *****  **      ******* **        **** |
| November '96  # 3 |                        **              **             |
+-------------------+-------------------------------------------------------+
| Newsletter of the Network of the European Course in Tropical Epidemiology |
+---------------------------------------------------------------------------+
Report from the 1996 Course in London
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
Twenty-seven students from Angola, Brazil, Gambia, Ghana, Georgia (CIS),
Holland, India, Japan, Kenya, Malawi, Nigeria, Sudan, Sweden, Tanzania,
Thailand, Tunisia, Uganda, United Kingdom, United States, Yemen, and
Zimbabwe came together in August for the Fifteenth European Course in
Tropical Epidemiology (ECTE) hosted by the London School of Hygiene
and Tropical Medicine.
Overall evaluation of the course was very good, with a mean score for the
participants general judgement being 7.3 out of 9.0. The computer sessions
were very popular - the new all day format worked extremely well. The
group-work sessions were also well received. As in previous years, the
more experienced statistics group was more difficult to please.
Changes for 1997
-=-=-=-=-=-=-=-=
On the basis of the course evaluation, the course steering committee
approved several changes to the course format that will be phased in over
the coming years.
More sessions will be organised as larger teaching blocks. The risk
sessions, which could easily take a whole day but are rushed in the
current timetable, were identified as a suitable module for this new
approach.
An applied statistical methods module will be available as a group-work
option. This should satisfy participants who would like to see some advanced
statistics in the course. The other group-work sessions will remain at the
introductory level but more options may be added (including maternal and
child health, sexual and reproductive health, and epidemiology in emergency
situations). The current double-session on nutritional surveillance
sessions will be dropped and replaced by a nutritional epidemiology group
work option. Sessions on health service research, epidemiology of chronic
disease, and a highly practical introduction to epidemiology session will
also be added.
Some of these changes will be seen in the 1997 course which will be hosted
by the Institute of Tropical Medicine in Heidelberg, Germany. For further
information about the 1997 course contact:
      Ms Anke Nitschke
      Department of Tropical Hygiene and Public Health
      Im Neuenheimer Feld 324
      D-69120 Heidelberg
      Germany
      Tel:        +49 6221 564905
      Fax:        +49 6221 564918
      E-mail:   anke.nitschke@urz.uni-heidelberg.de
The Future
-=-=-=-=-=
The future for the ECTE is bright. Student numbers remain at viable levels
(even at the traditionally low course fees). Course evaluations also remain
good. The search is on for a venue for the 1998, 1999, and 2000 courses.
Potential venues include Portugal, Holland, and Wales. Efforts will also be
made to find a suitable venue in Eastern Europe.
CDC announce a new version of EpiInfo
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
The latest version of the public domain medical statistics package EpiInfo
(v6.04a) has just been released by the Epidemiological Program Office of
the Centres for Disease Control. Copies of the program are available on the
Internet and from Brixton Books. EpiInfo is now supported by an
Internet mailing list. You can subscribe to this list or download
a copy of the latest version from:
      http://mkn.co.uk/help/extra/people/brixton_books
Work has started on a WindowsTM version of the program which should be
available by Summer 1997. More information will appear in TropEpi when a
review copy is available.
Personals
-=-=-=-=-
It is with deep regret that we report the death of Nicola Dollimore who
died in a caving accident this summer. Nicola was a motivating force for
the course and had taught and developed the statistics modules over many
years. She will be sadly missed by all who knew her.
ECTE 1997
-=-=-=-=-
The 16th ECTE will be held from the 8th - 20th September 1997 and will
be organised and hosted by the Institute of Tropical Medicine in
Heidelberg, Germany in collaboration with:
      DeptEpidemiology and Community Medicine Universtity of Antwerp, Belgium
      Institute of Tropical Medicine, Antwerp, Belgium
      National Institute for Public Health, Prague, Czech Republic
      Bureau for Topical Medicine and Hygiene, Copenhagen, Denmark
      London School of Hygiene and Tropical Medicine, London, UK
      GTZ GmbH, Eschborn, Germany
      Seccion de Medicina Tropical, Hospital Clinic, Barcelona, Spain
      Nordic School of Public Health, Göteborg, Sweden
      Swiss Tropical Institute, Basel, Switzerland
      Liverpool School of Tropical Medicine, Liverpool, UK
      Brixton Books / Brixton Health, Llanidloes, Wales
The European Course in Tropical Epidemiology
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
The European Course in Tropical Epidemiology is a collaborative venture
between various European institutes of tropical medicine and private sector
organisations. It is held annually in a different location.
This intensive introductory course in epidemiology is intended for
physicians, nurses, health programme managers, health administrators,
NGO field workers, and other persons with a professional interest in
health in tropical countries. The course provides participants with
basic skills in the epidemiological assessment of health problems and
service priorities and in the planning and execution of field studies.
Emphasis is placed on methodology, the practical application of
epidemiological tools in developing countries, the interpretation
of data, and the reporting of field studies.
This is an introductory level course and is appropriate for those with
no formal training in epidemiology or statistics. The course is very
intensive. Applicants should have a good command of English.
A certificate of attendance will be awarded to those completing the course.
Outline of the program
-=-=-=-=-=-=-=-=-=-=-=
Teaching will include lectures, seminars, practical exercises, and
group-work. Use will be made of the epidemiological software package
EpiInfo to illustrate and support teaching. The following subjects
will be included:
      Basic epidemiology of infectious diseases
      Basic epidemiology of non-infectious diseases
      Risk estimation, risk groups, risk factors, and risk markers
      Screening tests and screening programmes
      Epidemics and outbreaks
      Nutritional epidemiology
      Qualitative methods in epidemiology
      Sampling and sample size calculation
      Survey design and organisation
      Questionnaire design
      Data management
      Basic medical statistics using computers
      Textbooks and course materials
A course manual providing comprehensive study material will be given to
participants at the start of the course. For those interested in preparatory
reading the following texts are recommended:
      Epidemiology in Medicine, Hennekens CH, Buring JE, Little Brown, 1987
      Essentials of Medical Statistics, Kirkwood BR, Blackwell, 1988
      Course fee
The total course fee is DM2000. This includes participation in the course,
course material, incidental refreshments, and reception but does not include
travel costs, accommodation, or meals. The full course fee is payable on
notification of acceptance on the course. Admission will be confirmed when
payment is received. Since the number of places is limited, priority will
be given to early applicants. Applicants from developing countries are
advised to contact sponsoring organisations immediately.
Accommodation
-=-=-=-=-=-=-
Accommodation is available at a hall of residence close to the Department,
which offers self-catering facilities. The price will be approximately
DM250 for the two weeks and does not include breakfast. Lunch is available
from the University canteen (costing around DM5).
Participants should arrive in Heidelberg on Sunday 7th September. Heidelberg
is situated 70km south of Frankfurt (the nearest main international airport)
and can be reached in 1-2 hours by either train or bus.
Applying for a place on the Course
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
If you want to apply to attend the 16th European Course in Tropical
Epidemiology you should contcat:
      Ms Anke Nitschke
      Department of Tropical Hygiene and Public Health
      Im Neuenheimer Feld 324
      D-69120 Heidelberg
      Germany
      Tel:        +49 6221 564905
      Fax:        +49 6221 564918
      E-mail:   anke.nitschke@urz.uni-heidelberg.de

-- 
Mark Myatt

Dennis Roberts  writes:
>DOES A SIMPLE REGRESSION ANALYSIS ... REQUIRE A NORMALLY DISTRIBUTED
>DEPENDENT MEASURE? WHAT ABOUT THE INDEPENDENT VARIABLE?
>At 10:46 PM 12/17/96 EST, you wrote:
>>        We need to do regression analysis with RATIO (e.g., cost/benefit) as a
>>dependent variable, but I was told that I might violate normal distribution
>>assumption if I do so.  Any comments, suggestions?
Assume we are discussion ordinary least squares regression.
1.  Parameter estimates are least squares regardless of the parent
distribution.
2.  How useful those estimates are is another issue.
3.  In the absense of large samples, approximate normality
enables the distribution of the test statistics under the
null hypothesis to be known.
4.  "Large" enough samples obviate the need to know the
parent distribution.
5.  How large "large" enough is depends on how badly the
population is distributed.
6.  The randomness of the sampling and the independence
of the observations are of key importance.
-- 
--(Signature)      Robert M. Hamer hamer@rci.rutgers.edu 908 235 4218
  Do not send me unsolicited email advertisements.  I have never and
  will never buy.  I will complain to your postmaster.
  "Mit der Dummheit kaempfen Goetter selbst vergebens" -- Schiller

> it won't be long before the entire business of publishing scholarly wor=
ks
> ... will be drastically altered given the web and internet. the notion =
of
> reviewed submissions will be outmoded (for good or for bad) soon .... =
and
> papers etc. will merely be posted to the web ... for all to
> use/read/take/absorb ...
>
> the day of the printed/referred journal is rapidly limited ...
There will still be a place for reviewed "journals" or their equivalent =
on the web: otherwise we'd all
overwhelmed.  A reviewed electroinc journal will sift out the better work=
, and still have all the
advantages of the web (speed, links, animations etc.).
Looking forward to it.
-----------------------------------------------------------------
barryh@agb.com.au
Barry Haworth, Sampling Statistician, AGB McNair
PO Box 507, North Sydney 2059,         AUSTRALIA
            (02) 9911 7200 (ph)  (02) 9959 4947 (fax)
http://www.agb.com.au

On Mon, 16 Dec 1996 17:34:09 -0500, "Joseph K. Lyou" 
wrote:
>I want to analyze whether there is a significant trend over time in the
>annual failure rate of a product.  I have 20 years of measurements (i.e., n =
>20).  As I understand it, an ordinary regression analysis would be
>inappropriate because the residuals are not independent (i.e., the error
>associated with a failure rate for 1974 is more highly correlated with the
>1975 failure rate than the 1994 failure rate).  Is it appropriate to simply
>divide the data into two groups (the 1st 10 years vs. the 2nd 10 years) and
>do a between-groups ANOVA?  Or is there some other (better) way to analyze
>these data?
>
>Should anyone be so inclined as to do the analysis, here are the data:
>
>Year   Failure Rate
>1974   3.3
>1975   2.5
>1976   2.7
>1977   2.4
>1978   5.7
>1979   3.2
>1980   1.6
>1981   5.2
>1982   2.8
>1983   2.4
>1984   2.7
>1985   1.3
>1986   4.5
>1987   4.5
>1988   1.4
>1989   3.6
>1990   1.5
>1991   1.4
>1992   1.6
>1993   1.6
i would first of all try to apply the (qualitative) cox-stuart trend
test.
alex
========================================
DI Alexander Schatten
Institute for General Chemistry
University of Technology Vienna
email: aschatt@fbch.tuwien.ac.at
URL: http://echm10.tuwien.ac.at/inst/as
Tel.: +43 1 914-29-84
========================================

Dennis Roberts writes:
>it won't be long before the entire business of publishing scholarly works
>... will be drastically altered given the web and internet. the notion of
>reviewed submissions will be outmoded (for good or for bad) soon .... and
>papers etc. will merely be posted to the web ... for all to
>use/read/take/absorb ...
>
>the day of the printed/referred journal is rapidly limited ...
While dreamy eyed predictions of the future are fun, reality has a way
of ruining things.  Just like how I get a busy signal every time I try
to use my America Online account.
First, there is nothing inherent about web publications that prevent
the use of peer review, just like there is nothing inherent in paper
publications that mandate the use of peer review.
Second, Internet will no more replace paper publications any more than
television replaced movies and theater.  Both Internet and paper
publications are likely to co-exist for a long time.
Steve Simon, ssimon@cmh.edu, Standard Disclaimer.
-------------------==== Posted via Deja News ====-----------------------
      http://www.dejanews.com/     Search, Read, Post to Usenet

                                   *
                                 * * *
       Peace                   * * * * *
                                * * * *
         to                   * * * * * *
                            * * * * * * * *
    One and All,              * * * * * * *
                           * * * * * * * * *
   Now and in the            * * * * * * *
                              * * * * * *
     New Year!!          * * * * * * * * * * *
                             * * * * * * *
                                 * * *
                                 * * *
 WBW 1996
______________________________________________________________________________
William B. Ware, Professor and Chair                   Educational Psychology
CB# 3500                                               EMAIL:   wbware@unc.edu
University of North Carolina                           PHONE:   (919)-966-5266
Chapel Hill, NC      27599-3500                        FAX:     (919)-962-1533
URL:http://www.unc.edu/~wbware/
______________________________________________________________________________

Hi,
If anyone is researching on material like PAR models and periodic
integration and cointegration, I would appreciate some help with
certain problems I'm having.
Please reply by email.
Many thanks.
Kenneth.

I've been exposed to the three Erlang models (A, B or Poisson, C) where
blocked telephone calls are cleared, delayed or held.
Now I'm looking at modeling tasks of a more complex nature such as
writing 100 C++ subroutines or building 100 regression models.  If each
regression model averaged 3 hours to build, I could use the Poisson
distribution to find out how many data analysts I would need to ensure
only one chance in 20 that a model building request would be "blocked",
if we experience a demand of 5 model building requests per day.
BUT setup of telephone calls is usually quite quick and almost NEVER
interrupted, while a task which averages 3 hours will almost ALWAWS be
interrupted, at least once, so --
QUESTION: is there a model which accounts for task interruptions, or should
I just ignore them and use the Poisson?
The problem is complicated by the fact that some interruptions (like LUNCH) make
the analyst unavailable, while others (need for a corrected dataset) actually
make the analyst available to build other models.
Just think -- if there's a solution to this, we'll be able to schedule data
analysts for split shifts just like we now do with telephone operators or 
counter people at McDonalds -- HEY -- there could be a: 
"MODELS-R-US - Regression Models while you Wait - Your model in 3 hours or IT'S FREE"
Well Howell (whowell@superlink.net)

In article
,  Barry
Haworth  writes:
|> > it won't  be  long  before  the  entire  business  of  publishing
|> > scholarly works ... will be drastically altered given the web and
|> > internet. the notion of reviewed  submissions  will  be  outmoded
|> > (for good or for bad) soon .... and papers etc.  will  merely  be
|> > posted to the web ... for all to use/read/take/absorb ...
|> >
|> > the day of the printed/referred journal is rapidly limited ...
|>
|> There will still be  a  place  for  reviewed  "journals"  or  their
|> equivalent on the web: otherwise we'd all overwhelmed.  A  reviewed
|> electroinc journal will sift out the better work,  and  still  have
|> all the advantages of the web (speed, links, animations etc.).
Why not just let links take the place of reviewing?  After  all,  what
is a review except the sanction  of  a  recognized  authority?   If  I
recognized you as an authority on statistics, I'd just go to your home
page and examine the links to papers you'd read recently and  approved
of.
|> Looking forward to it.
Ditto.
--
Randy Tobias          SAS Institute Inc.     sasrdt@unx.sas.com
(919) 677-8000 x7933  SAS Campus Dr.         us024621@interramp.com
(919) 677-8123 (Fax)  Cary, NC   27513-2414
   Faith, faith is an island in the setting sun.
   But proof, yes: proof is the bottom line for everyone.
                                                       -- Paul Simon

makes sense to me but ... the powers to be for promotion and tenure seem to
want MORE than that. personally ... since the appearance in a good journal
does not assure that the article/study was good ... i prefer to make my work
available .. and let those who access it decide if it is useful/good or not.
the acceptance by a journal does not make it so ... but we have come to rely
SO heavily on its appearance in a decent journal being THE proof of worth
... i think we have forgotten that intelligent readers can figure that out
on their own ..
At 04:00 PM 12/19/96 GMT, you wrote:
>In article
>,  Barry
>Haworth  writes:
>|> > it won't  be  long  before  the  entire  business  of  publishing
>|> > scholarly works ... will be drastically altered given the web and
>|> > internet. the notion of reviewed  submissions  will  be  outmoded
>|> > (for good or for bad) soon .... and papers etc.  will  merely  be
>|> > posted to the web ... for all to use/read/take/absorb ...
>|> >
>|> > the day of the printed/referred journal is rapidly limited ...
>|>
>|> There will still be  a  place  for  reviewed  "journals"  or  their
>|> equivalent on the web: otherwise we'd all overwhelmed.  A  reviewed
>|> electroinc journal will sift out the better work,  and  still  have
>|> all the advantages of the web (speed, links, animations etc.).
>
>Why not just let links take the place of reviewing?  After  all,  what
>is a review except the sanction  of  a  recognized  authority?   If  I
>recognized you as an authority on statistics, I'd just go to your home
>page and examine the links to papers you'd read recently and  approved
>of.
>
>|> Looking forward to it.
>
>Ditto.
>
>
>--
>Randy Tobias          SAS Institute Inc.     sasrdt@unx.sas.com
>(919) 677-8000 x7933  SAS Campus Dr.         us024621@interramp.com
>(919) 677-8123 (Fax)  Cary, NC   27513-2414
>
>   Faith, faith is an island in the setting sun.
>   But proof, yes: proof is the bottom line for everyone.
>                                                       -- Paul Simon
>
>
===========================
 Dennis Roberts, Professor EdPsy             !!! GO NITTANY LIONS !!!
 208 Cedar, Penn State, University Park, PA 16802 AC 814-863-2401
 WEB (personal) http://www2.ed.psu.edu/espse/staff/droberts/drober~1.htm

D.C.Lee (DLEE@cms.cc.wayne.edu) wrote:
: Hello All,
:  
: I am seeking to develop a predictive model of a particular
: phisiological parameter(response variable) using to other
: phisiological parameters as predictor variables. Of the
: response variable, or parameter, I am interested in events
: which occur only  0 to 5 times during a four hour test run.The
: event lasts only five or ten minutes as a spike and has a
: relatively quick decay. The response variable is nearly
  -- 5 minutes plus the surrounding hour, taken 5 times, is
more than the 4-hour test.
But if you have a bunch of 0-0-0  sort of data, it will contribute
nothing to the useful prediction, once you have accounted for
a LITTLE of it.  It might make your R-squared look bigger, by 
predicting a lot more zero-scores, even though the error (computed
residuals) are getting bigger for the peaks that you are really
interested in.
If you are trying to *predict*  a "spike", is there ANY purpose in
looking at the data for the next half hour?  Do your spikes ever
double up that way?
Deciding on transformations is something that you do while looking at
the data with multiple spikes for a person.  Does a transformation
preserve (or even, enhance) the similarities of the spikes for one
person?  If the data for one person looks 'typical' for that PERSON,
then the proper analysis would create scores or parameters of some
kind, by PERSON, and then do your eventual testing across N=20
people.  
You might do an analysis controlling for Subject, but I think you 
would never drop the dummy-variables for Subject.  That seems too 
much like double-counting your sample. 
Rich Ulrich, biostatistician              wpilib+@pitt.edu
Western Psychiatric Inst. and Clinic   Univ. of Pittsburgh

Peter Midford (pmidford@students.wisc.edu) wrote:
:        I'm wondering if it is possible to calculate power values (1 - Beta)
: for one or all of the common contingency table tests (e.g. chi-square, G,
: Fisher's exact).
  -- Most any text, for Pearson's contingency table chisquare.  Try
Cohen, "Statistical Power Analysis for the Behavioral Sciences".
Rich Ulrich, wpilib+@pitt.edu

Many thanks for everybody's input.  It seems a statistician is a
necessity (as I expected) and I'll have to wait.  
I'm sorry I glossed over the study's structure, as I underestimated the
medical knowledge present in this group.  All patients in the study were
treated by 8 doctors of the Helicopter Emergency Medical service working
to strict guidelines, and most details of the prehospital phase are
available for analysis.  RTS, and other physiological scores are taken
from the first readings made on scene (around 10 minutes into the
incident), and so should be unaffected by time spent on scene).  All
medical interventions have been recorded, along with details of
extrication etc.
Basically a rough over view of the data seems to suggest that time
spent  on scene is not correlated with survival, ITU stay, Hospital Stay
or functional independence at 12 months.  But obviously a meticulous
statistical analysis will need to be performed to confirm or deny this.
Once again many thanks for your input, and I am impressed by the
knowledge displayed.
-- 
__________________________________________________
Dr. Karim Brohi BSc FRCS FRCA
Trauma & Critical Care Unit, Royal London Hospital
Mailto:karim@trauma.org
http://www.trauma.org/

We are testing a new lab test versus the standard test.  We will calculate
sensitivity and specificity of the new test.  How do I determine the sample
size needed to have valid results?
Susan Adlis
adliss@found.hsmnet.com

what do you mean by "valid"???
if you are asking ... to find out if there is really a true difference in
effect ... but, we know of course that the two cannot have the same
IDENTICAL impact. so ...
you must try to rationally decide on what SIZE of an impact you would
consider to be important ... AT A MINIUM.
the question you have asked on the surface would seem to have a simple
answer but .. it does not.
At 01:46 PM 12/19/96 PST, you wrote:
>We are testing a new lab test versus the standard test.  We will calculate
>sensitivity and specificity of the new test.  How do I determine the sample
>size needed to have valid results?
>
>Susan Adlis
>adliss@found.hsmnet.com
>
>
===========================
 Dennis Roberts, Professor EdPsy             !!! GO NITTANY LIONS !!!
 208 Cedar, Penn State, University Park, PA 16802 AC 814-863-2401
 WEB (personal) http://www2.ed.psu.edu/espse/staff/droberts/drober~1.htm

nakhob@mat.ulaval.ca (Renaud Langis) writes:
>On Fri, 13 Dec 1996 00:07:22 -0500, Ya-Fen Lo 
>wrote:
>>Is it possible to perform tests of simple effects
>>(as defined in APPLIED STATISTICS by HINKEL/WIERSMA/JURS)
>>in SAS ? I am using the following setup
Yes.
>You can use the TEST statement in proc GLM. May be also in proc ANOVA. Do you
>simply want to know if an effect is significant? if so, just check the ANOVA
>table.
That is not what the original question asked.  That person wants
to contrast levels of one effect at specific levels of the other
effect.  One has to do that with CONTRAST or ESTIMATE statements.
>I suppose this is just a typing error but CLASSES should be written CLASS.
Actually, CLASS works just fine.  It is one of the several
alternative forms of the statement available.
-- 
--(Signature)      Robert M. Hamer hamer@rci.rutgers.edu 908 235 4218
  Do not send me unsolicited email advertisements.  I have never and
  will never buy.  I will complain to your postmaster.
  "Mit der Dummheit kaempfen Goetter selbst vergebens" -- Schiller

Richard F Ulrich (wpilib+@pitt.edu) wrote:
: Barry Haworth (barryh@AGB.COM.AU) wrote:
:  << concerning the sampling of 300 out of 3000 ... >>
: 
: :" Taking a sample of a sample is perfectly appropriate, so long as
: the second sample is drawn in a sensible way (a random sample of all
: the original responses, for example)"
(The 3 lines immediately above are from my reply.)
: 
:   -- Please, if you have to analyze the 300, then draw them 
: SYSTEMATICALLY rather than randomly.  For instance, it is VERY OFTEN
: useful to know if the early (fast?)  respondents were different
: from those whose forms came in last.  (In mortality followups, 
: one looks to see if the causes of death are different for those
: whose records were hardest to find  -  found by 'only one method'
: or found only by persistent checking.)  
: 
In the survey literature, systematic sampling refers to taking every k-th
observation (say, every 10th one in this case).  Provided that the first
one is randomly selected (random start), this is a form of random or
probability sampling (not simple random sampling).  I was really trying to
guard against taking a "convenience" sample  -- these can lead to big
trouble.  I think we have here a difference in terminology more than
anything else.
: You might draw 100 early+middle+late for your 300, so you could
: compare.  Big changes across the strata suggest a big chance 
: that your Unsampled cases are even more different than early vs late.
: 
Fine.  A stratified random sample would do nicely.
-- 
Michael P. Cohen                       home phone   202-232-4651
1615 Q Street NW #T-1                  office phone 202-219-1917
Washington, DC 20009-6331              office fax   202-219-2061
mcohen@cpcug.org

************* 1997 Course Announcement *********
      MODERN REGRESSION AND CLASSIFICATION
       Waikiki, Hawaii: February 17-18, 1997 
*************************************************
A two-day course on widely applicable statistical methods for
modeling and prediction, featuring
Professor Trevor Hastie    and   Professor Robert Tibshirani
Stanford University              University of Toronto
This course was offered and enthusiastically attended at five
different locations in the USA in 1996.
This two day course covers modern tools for statistical prediction and
classification. We start from square one, with a review of linear
techniques for regression and classification, and then take attendees
through a tour of:
 o  Flexible regression techniques
 o  Classification and regression trees
 o  Neural networks
 o  Projection pursuit regression
 o  Nearest Neighbor methods
 o  Learning vector quantization
 o  Wavelets
 o  Bootstrap and cross-validation
We will also illustrate software tools for implementing the methods.
Our objective is to provide attendees with the background and
knowledge necessary to apply these modern tools to solve their own
real-world problems. The course is geared for:
     o  Statisticians
     o  Financial analysts
     o  Industrial managers 
     o  Medical and Quantitative  researchers
     o  Scientists
     o  others interested in  prediction and  classification
Attendees should have an undergraduate degree in a quantitative
field, or have knowledge and experience working in such a field.
PRICE: $750 per attendee if received by January 15, 1997. Full time
registered students receive a 40% discount.  Attendance is limited to
the first 60 applicants, so sign up soon!  These courses fill up
quickly.
TO REGISTER: Fill in and return the form appended.
For more details on the course and the instructors:
   o point your web browser to: 
        http://stat.stanford.edu/~trevor/mrc.html
        OR send a request by
   o FAX to Prof. T. Hastie at (415) 326-0854, OR
   o email to trevor@stat.stanford.edu
<----------------------------- Cut Here ------------------------------->
 Please print, and fill in the hard copy to return by mail or FAX
                                REGISTRATION FORM
                    Modern Regression and Classification
             Monday, February 17 and Tuesday, February 18, 1997.
          Hilton Hawaiian Village, Waikiki Beach, Honolulu, Hawaii.
         Name   ___________________________________________________
                Last                 First                   Middle
         Firm or Institution  ______________________________________
        Standard Registration ____         Student Registration ____
         Mailing Address (for receipt)     _________________________
         __________________________________________________________
         __________________________________________________________
         __________________________________________________________
          Country                    Phone                      FAX
         __________________________________________________________
                               email address
       __________________________________________     _______________
       Credit card # (if payment by credit card)      Expiration Date
                  (Lunch preference - tick as appropriate):
         ___ Vegetarian                           ___ Non-Vegetarian
Fee payment can be made by MONEY ORDER , PERSONAL CHECK, or CREDIT CARD
(Mastercard or Visa.) For checks and money orders: all amounts are given in
US dollar figures. Make fee payable to Prof. T. Hastie. Mail it, together
with this completed Registration Form to:
Prof. T. Hastie
538 Campus Drive
Stanford
CA 94305
USA
For payment by credit card, include credit card details above, and mail to
above address, or else FAX form to 415-326-0854
For further information, contact:
Trevor Hastie
Stanford University
Tel. or FAX: 415-326-0854
e-mail: trevor@stat.stanford.edu.
http://stat.stanford.edu/~trevor/mrc.html
REGISTRATION FEE
Standard Registration: U.S. $750 ($950 after Jan 15, 1997)
Student Registration: U.S. $450 ($530 after Jan 15, 1997)
Student registrations - include copy of student ID.
- Cancellation policy: No fee if cancellation before Jan 15, 1997.
- Cancellation fee after January 15 but before Feb 12, 1997: $100. 
- Refund at discretion of organizers if cancellation after Feb 12, 1997.
- Registration fee includes course materials, coffee breaks, and lunches
- On-site Registration is possible if course is not fully booked, at late
fee.

Isn't this a path analysis question?  The severity of the injury determines
both the patients' health and the amount of time spent at the scence (with
more severe injuries requiring a greater amount of time).  In addition, the
amount of time spent at the scene may have some effect on the patients'
health.  The researcher appear primarily interested in this path, especially
whether the relationship is positive or negative.
     Severity of Injury
             |          \
             |           \
             |            \
            \/            \/
     Time on Scene -------> Patient Health Outcome

Here are some sources:
BTW: Be careful with CHAID...it will usually find differences. Continue to test variables
it finds by using a logistic regression model in the case of a categorical dependent
variable and correlations and Anova for an interval or ratio level dependent variable.
I usually split my sample into one or more parts too.  One to develop the CHAID tree,
and at least one other to test prediction accuracy on.  
Some early sources, though not probably what you are looking for.  They suggest a
general framework for classifcation analysis.
  W.A. Belson, "A technique for studying the effects of television broadcast", Applied
  Statistics, 5, 195 (1956).
  Lazarsfield, Paul, F. and Rosenberg, Morris (editors) The Lanugage of Social
  Research, A Reader in the Methodology of Social Research.  Pg 118-120.
  Sonnquest, J.A. and Morgan, J.N. "The detection of interaction effects" (1964)  Ann
  Arbor:  Institute for Social Research, University of Michigan.
  -  This resulted in the AID algorithm.
Several papers by Leo Goodman.  Starting with...
  Goodman, Leo, A. "On partioning chi squared and detecting partial association in three-
  way contingency tables."  Journal of the Royal Statistical Society, Series B, 31. (1969)
  Goodman, Leo, A. ""The mulivariate analysis of qualitative data: interactions among
  multiple classifications." Journal of the American Statistical Association, 65, 226-56.
  -  Reprinted in his book (1978) Analyzing Qualitative/Categorical Data, edited by J.
     Magidson, Cambridge, Massachusetts: Abt Books.
  Goodman, Leo, A. "The analysis of cross-classified data having ordered and/or unordered
  categories: association models, correlation models, and asymetry models for contingency
  tables with or without missing entries."  The Annals of Statistics, 13(1), 10-69. (1985)
  -  This was the start of Jay Magidson's enhancement to the CHAID model (see the last
     reference below) now implemented in SPSS's SI-CHAID and an option in SAS.  I use
     KnowledgeSEEKER, but this enhancement has been made there.
  Goodman, Leo. A. "Measures, models and graphic displays in the analysis of cross-classified
  data."  Journal of the American Statistical Association, 86, 1085-1138. (1991).
Leo Brieman, J.H. Frieman, R.A. Olshen and C.J. Stone, Classification and Regression
Trees (Wadsworth, 1984)
Biggs, David, with Barry DeVille & Ed Suen. "A method of choosing multiway partitions
for classification and decision trees" in the Journal of Applied Statistics, Vol 18,
No. 1, )1991).
  - This is KnowledgeSEEKER's implementation of CHAID.
G.V. Kass, "An exploratory technique for investigating large quantities of categorical
data", Applied Statistics, 29, 2, 1980, 119-127. and really the same thing in
D.M.Hawkins and G.V. Kass, "Automatic interaction detection", in D.G. Hawkins (ed.),
Topics in Applied Multivariate Analysis (Cambridge, Cambridge University Press, 1982)
-  This was the basis for CHAID.
Magidson, Jay, "Chi-squared analysis of a scalable dependent variable", Proceedings of
the 1992 annual meeting of the American Statistical Association, Section on Statistical
Education. (1992)

In article <59ctcv$43g@news5.wavefront.com>, diffsimilar@wavefront.com says...
I just read my own post.  Jay Magidson's enhancement
to the CHAID model, now implemented in SPSS's SI-CHAID has
NOT been implemented in KnowledgeSEEKER.  I typed the list
too fast.  Sorry 'bout that.

The article you are looking for as a reference to CHAID can be found in
the Journal of Marketing Research (JMR).  I believe the year was 1978 and
one of the two authors is Perreault.  However, Perreault and his co-author
reference a dissertation by Kass as the original reference.
CHAID can also be purchased as a stand-along module from Statistical
Innovations, Inc in Belmont, Mass.  The principal of that company is Jay
Magidson, who re-wrote Perreault's code and eventually licensed it to
SPSS.  He has written several articles in both JMR and the Journal of
Direct Marketing on the use of CHAID, all very good reading.
Steve Cohen
Stratford Associates
617-928-0110 x104
StratfordA@aol.com

Hi there....would appreciate some assistance with the following: I am
analyzing a data base with 153 in a treatment group and approximately 350 in
the control group; the DV is cost of medical services with the procedure
involving coronary bypass......the tx group may have only 3 or 4 patients
with this procedure with possibly slightly more in the control
group.....thus, the data points are predominantly 0 (zero).....obviously, any
of the parametric statistics are inappropriate.....a colleague indicated that
he recently purchased STATA which is able to analyze Poisson or Tobit
distributions, which may suit my data....this is foreign territory for
me......any suggestions how to compare the two groups when data points are
primarily zero....the means make no sense (e.g., average of $200 for tx
group).........is there anything in SPSS which is appropriate (e.g., using
COMPUTE function to create Poisson distribution)..
I am truly befuddled!!!....thank you and have a great holiday
Dale Glaser
Clinical Research; Sharp HealthCare
dale.glaser@sharp.com
daleglas@aol.com

In article <26222628@vixen.Dartmouth.EDU>, Haiyi Xie 
writes:
|>         We need to do regression analysis with RATIO (e.g., cost/benefit) as
a
|> dependent variable, but I was told that I might violate normal distribution
|> assumption if I do so.  Any comments, suggestions?
Normality is not the main issue. When you compare two cost/benefit
ratios, do you look at the difference of the ratios or the ratio
of the ratios? I would guess the latter, in which case you should
use the log of the cost/benefit ratio as the dependent variable.
--
Warren S. Sarle       SAS Institute Inc.   The opinions expressed here
saswss@unx.sas.com    SAS Campus Drive     are mine and not necessarily
(919) 677-8000        Cary, NC 27513, USA  those of SAS Institute.
 *** Do not send me unsolicited commercial or political email! ***

>Isn't this a path analysis question?  The severity of the injury determines
>both the patients' health and the amount of time spent at the scence (with
>more severe injuries requiring a greater amount of time).  In addition, the
>amount of time spent at the scene may have some effect on the patients'
>health.  The researcher appear primarily interested in this path, especially
>whether the relationship is positive or negative.
>
>     Severity of Injury
>             |          \
>             |           \
>             |            \
>            \/            \/
>     Time on Scene -------> Patient Health Outcome
>
Yes, but consider the problem of measurement error. Severity of injury is
correlated with time on scene, but severity is measured with much more
error than time. Thus, time will have a correlation with patient health
outcome because it contains information on severity of injury that is not
in the measures designed to measure severity. Imagine two cases where the
severity of injury scores were the same but the crew spent more time on
the scene in one case. This is likely to reflect unmeasured aspects of
severity as well as variations in time on scene not due to severity. This
will produce a spurious correlation.
The more I think about it, the less I can imagine a model that would be
an appropriate mapping of the process being studied onto a statistical
procedure.
_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/
    _/_/_/      _/_/     _/_/_/     _/     Ronan M Conroy
   _/    _/   _/   _/  _/          _/      Lecturer in Biostatistics
  _/_/_/    _/          _/_/_/    _/       Royal College of Surgeons
 _/   _/     _/              _/  _/        Dublin 2, Ireland
_/     _/     _/_/     _/_/_/   _/         voice +353 1 402 2431
                                           fax   +353 1 402 2329
_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/
'Do not try to be a genius in every bar' [Brendan Behan? - No, Faure!]