[ABE-L] rede da ASA
Jose Carvalho
carvalho em statistika.com.br
Qua Nov 4 10:02:15 -03 2015
Chega a ser irritante. Pessoas de outras áreas "redescobrem" assuntos já
velhos na literatura estatística, renomeiam as técnicas e conceitos e
saem falando como se fosse coisa nova. Esta análise de redundância
parece mais um caso desses. A área médica está repleta de
"contribuições" em experimentação e análise estatística (mormente de
famosos ingleses - da área médica) que se constituem em desrespeito ao
passado. Claro que se deve fazer uma revisão criteriosa da literatura e
pagar-se o tributo àqueles que tem precedência. Gente, não vou citar
exemplos, pois arranjaria briga e talvez perdesse trabalhos em
organismos internacionais, que estão sob fortíssima influência de
"usurpadores".
Pedro é uma pessoa gentil. Acho foi, de leve, apenas irônico. Mas o
assunto é parte de algo sério. Quem trabalha em ensaio clínico tem de
conviver com termos novos para coisa velhas... Em ensaios clínicos, há
um agravante: as traduções ao português tem exemplos risíveis. O mesmo
ocorre em testes com consumidores, onde se confunde um esquema de
observação (como obter o dado do voluntário) com o esquema experimental;
daí, erra-se frequentemente na estimativa dos erros e precisões de
estimativas. Seria preciso que essa turma toda tratasse de estudar,
antes de se meter na prática.
Dentro em pouco regressão multivariada ou análise canônica será chamada
de análise de redundância. E será uma "nova" técnica de algum psicometri
sta.
Abraços a todos.
Zé Carvalho
On 11/04/2015 09:09 AM, pam em ime.usp.br wrote:
> Caros leitores:
> A American Statistical Association (ASA) não mantinha rede para
> eventuais comunicações entre associados, como a ABE tem. Provavelmente
> porque com cerca de 15000 associados, a coisa poderia facilmente virar
> bagunça.
>
> Recentemente foi criada o que se chama "ASA Connect Digest", e abaixo
> reproduzo uma edição de 3/11/2015.
>
> Os assuntos discutidos são de natureza estatística somente. Por esta
> última fiquei sabendo que existe algo chamado "redundancy analysis"!
>
> Pedro
>
> =====================================
>
> Please Do Not Forward, use the options to the right to forward or rep
ly.
>
>
> Email digest for the ASA Connect egroup.
> ----------------------------------------------------------------------
--------------------------------------
>
>
> 1. RE: What's your Twitter handle?
>
> 2. RE: Principal Components Analysis
>
> 3. RE: What's your Twitter handle?
>
> 4. RE: Principal Components Analysis
>
> 5. RE: Principal Components Analysis
>
> 6. RE: Principal Components Analysis
>
> 7. RE: Principal Components Analysis
>
> 8. Tenure track faculty position at Rice University
>
> 9. RE: Principal Components Analysis
>
> ----------------------------------------------------------------------
--------------------------------------
>
>
> 1.From: Michelle Wiest
> Posted: Tuesday November 3, 2015 9:15 AM
> Subject: RE: What's your Twitter handle?
> Message:
> @MichelleWiest ...I'm a statistician/epidemiologist at the University
of
> Idaho. Happy to connect!
> ------------------------------
> Michelle Wiest
> Assistant Professor
> University of Idaho
> ----------------------------------------------------------------------
---
> Original Message:
> Sent: 11-02-2015 18:09
> From: Lara Harmon
> Subject: What's your Twitter handle?
>
>
> Are you on Twitter? What's your handle? No pressure to share, of cours
e,
> but if you'd like to--please do!
>
>
> (I'm @Amstat_Lara. Other ASA-related accounts include @Amstatnews,
> @ASA_SciPol, @Ron_Wasserstein, @Amstat_Meg, @Amstat_Sara, @Amstat_Mega
n,
> and @AmyFarrisASA. Feel free to follow any and/or all of us!)
> ------------------------------
> Lara Harmon
> Marketing and Online Community Coordinator
> American Statistical Association
> ------------------------------
>
>
>
>
>
>
> 2.From: Simon Blomberg
> Posted: Tuesday November 3, 2015 9:15 AM
> Subject: RE: Principal Components Analysis
> Message:
> I would like to second Pieter Kroonenberg's suggestion. Redundancy
> analysis is the multivariate version of regression, where you want to
> regress a set of response variables onto a set of predictor variables.
> This seems to be what you want? An alternative would be canonical
> correlation analysis, if you are just interested in the correlations,
> rather than the regression coefficients.
>
>
> Cheers,
>
>
>
> Simon.
> ------------------------------
> Simon Blomberg
> Lecturer and Consultant Statistician
> University of Queensland School of Biological Sciences
> ----------------------------------------------------------------------
---
> Original Message:
> Sent: 11-02-2015 01:53
> From: Pieter Kroonenberg
> Subject: Principal Components Analysis
>
>
> Dear Ms. Landon,
>
>
>
>
>
> What you would like to do is called redundancy analysis, it is a speci
al
> case of canonical correlation analysis. In the latter you look for
> linear combinations of each set of variables such that the correlation
> among the linear combinations from both set is high as possible, in th
e
> former case the criterion is that the linear combination of the respon
se
> set can be best predicted by the linear combination of the predictor
> set. Van den Wollenberg in Psychometrika deals with redundancy analysi
s.
> The latter seems to be due to Hotelling. A net search will give all th
e
> relevant references.
> ------------------------------
> Pieter Kroonenberg
> Leiden University
>
>
>
>
>
>
>
> 3.From: Peter Flom
> Posted: Tuesday November 3, 2015 9:15 AM
> Subject: RE: What's your Twitter handle?
> Message:
> I am on Twitter. My handle is @PeterFlomStat
>
>
>
> Peter
> ------------------------------
> Peter Flom
> Peter Flom Consulting
> ----------------------------------------------------------------------
---
> Original Message:
> Sent: 11-02-2015 18:09
> From: Lara Harmon
> Subject: What's your Twitter handle?
>
>
> Are you on Twitter? What's your handle? No pressure to share, of cours
e,
> but if you'd like to--please do!
>
>
> (I'm @Amstat_Lara. Other ASA-related accounts include @Amstatnews,
> @ASA_SciPol, @Ron_Wasserstein, @Amstat_Meg, @Amstat_Sara, @Amstat_Mega
n,
> and @AmyFarrisASA. Feel free to follow any and/or all of us!)
> ------------------------------
> Lara Harmon
> Marketing and Online Community Coordinator
> American Statistical Association
> ------------------------------
>
>
>
>
>
>
> 4.From: R. Cook
> Posted: Tuesday November 3, 2015 9:15 AM
> Subject: RE: Principal Components Analysis
> Message:
> I generally stay away from principal components in such situations
> because of method can miss relevant relationships. Partial least
> squares is not uniquely defined when regressing multiple responses on
> multiple predictors, as there are different algorithms that produce
> different answers. Canonical correlation analysis could be useful. A
n
> appropriate choice would seem to depend on the reason(s) for needing t
o
> use dimension reduction in the first place. For instance, additional
> issues arise if it's needed to compensate for a small sample size.
>
> ------------------------------
> R. Cook
> University of Minnesota
> ----------------------------------------------------------------------
---
> Original Message:
> Sent: 10-30-2015 09:27
> From: Linda Landon
> Subject: Principal Components Analysis
>
> Everyone,
>
> I'm about to make my first foray into principal component analysis sin
ce
> graduate school. I'm working on a multivariate regression problem tha
t
> has multiple independent variables and multiple dependent variables.
> I'm wondering if it would be possible to perform multivariate regressi
on
> by using BOTH the principal components representing dimensions in the
> set of independent variables as independent variables and the principa
l
> components representing dimensions in the set of dependent variables
as
> dependent variables in the same regression model. However, I'm unsure
> if this application is statistically valid.
>
> The common examples found in texts and in the peer-reviewed literature
> when PCA is applied prior to multivariate regression are
>
> (1) multiple independent variables to find the principal components
> in the independent variables and then one or more components are used
as
> independent variables in a regression against a single dependent
> variable or
> (2) to multiple dependent variables to find the principal component
s
> in the dependent variables and then a single component is used as a
> dependent variables in a regression against a more or more independent
> variables.
>
> However, I would like to know if it is possible to perform one
> multivariate regression analysis on two sets of principal components,
> one set of components representing the dimensions in the independent
> variables and the other set of components representing the dimensions
in
> the dependent variables. In my rather limited literature search, I
> haven't found an example of this application in the peer-reviewed
> literature, which could indicate that it is not statistically valid.
> Further, other sources of information don't explicitly address this
> application.
>
> Thanks in advance for your insight.
>
> Linda
>
> Linda A. Landon, PhD, ELS
> President
> Research Communiqu
> Jefferson City, MO
> Email: LandonPhD em ResearchCommunique.com <LandonPhD em ResearchCommunique
.com>
> Phone: 573-797-4517
> Central Standard Time ( CST ) = GMT-6 (November February)
> Central Daylight Time ( CDT ) = GMT-5 (March October)
>
>
>
>
>
>
>
>
> 5.From: Larry Price
> Posted: Tuesday November 3, 2015 9:15 AM
> Subject: RE: Principal Components Analysis
> Message:
> I agree with Ken Burnham, canonical correlation analysis will allow yo
u
> to answer your research questions. Canonical correlation is very
> flexible yet underutilized.
>
>
> Larry Price
>
>
> Texas State University
> ------------------------------
> Larry Price
> Director/Professor - Interdisciplinary Initiative for Research
> Texas State University
> ----------------------------------------------------------------------
---
> Original Message:
> Sent: 11-02-2015 01:04
> From: Kenneth Burnham
> Subject: Principal Components Analysis
>
>
> I think canonical correlation is the method you want.
> ------------------------------
> Kenneth Burnham
> Colorado State University
>
>
>
>
>
>
>
> 6.From: Thaddeus Tarpey
> Posted: Tuesday November 3, 2015 10:09 AM
> Subject: RE: Principal Components Analysis
> Message:
> This is a multivariate multiple regression problem. An approach to th
is
> problem that might be useful and insightful is to perform a reduced ra
nk
> regression. The book by Reinsel and Velu might be helpful for this
> approach.
>
>
> Thad Tarpey
> ------------------------------
> Thaddeus Tarpey
> Wright State University
> ----------------------------------------------------------------------
---
> Original Message:
> Sent: 11-02-2015 01:53
> From: Pieter Kroonenberg
> Subject: Principal Components Analysis
>
>
> Dear Ms. Landon,
>
>
>
>
>
> What you would like to do is called redundancy analysis, it is a speci
al
> case of canonical correlation analysis. In the latter you look for
> linear combinations of each set of variables such that the correlation
> among the linear combinations from both set is high as possible, in th
e
> former case the criterion is that the linear combination of the respon
se
> set can be best predicted by the linear combination of the predictor
> set. Van den Wollenberg in Psychometrika deals with redundancy analysi
s.
> The latter seems to be due to Hotelling. A net search will give all th
e
> relevant references.
> ------------------------------
> Pieter Kroonenberg
> Leiden University
>
>
>
>
>
>
>
> 7.From: Donald Myers
> Posted: Tuesday November 3, 2015 1:37 PM
> Subject: RE: Principal Components Analysis
> Message: There is also also an old paper in SIAM Review on Linear
> Dependency
> Analysis and a FORTRAN code
>
> Donald E Myers
>
> ------Original Message------
>
>
> Dear Dr. Landon,
>
>
> I would recommend reading up on CANONICAL CORRELATION analysis.
> This is essentially a technique that simultaneously explores & tests t
he
> relations of a SET of variables (e.g., y1, y2, y3, etc.) versus a SET
of
> variables (e.g., x1, x2, x3, etc.). For example, check the internet f
or
> SAS stats documentation re the CANCORR procedure ("Proc Cancorr").
>
>
> Joseph J. Locascio, Ph.D.
> ------------------------------
> Joseph J. Locascio, Ph.D.,
> Assistant Professor of Neurology,
> Harvard Medical School,
> and Statistician,
> Memory and Movement Disorders Units,
> Massachusetts Alzheimer's Disease Research Center,
> Neurology Dept.,
> Massachusetts General Hospital (MGH),
> Boston, Massachusetts 02114
> Phone: (617) 724-7192
> Email: JLocascio em partners.org
> ------------------------------
>
> 8.From: Marina Vannucci
> Posted: Tuesday November 3, 2015 4:22 PM
> Subject: Tenure track faculty position at Rice University
> Message:
>
> Please share this announcements with interested parties.
>
>
> The Department of Statistics at Rice University invites applications f
or
> a tenure-track position. Priority hiring is at the Assistant professor
> level, however, highly qualified, experienced candidates may be
> considered at the rank of Associate Professor. A doctorate in statisti
cs
> or a related field is required, with potential or proven excellence in
> research and teaching.
>
>
> Applications are sought from areas of modern statistics that will
> enhance the George R. Brown School of Engineerings focus on Data Scien
ce
> and the understanding of complex data, while strengthening the core
> research of the department. Example of areas include but are not limit
ed
> to Bayesian methods, computational finance, functional data,
> multivariate analysis, networks or graphical models, probability theor
y,
> statistical machine learning, spatial and temporal processes,
> statistical computing, stochastic processes and optimization. Several
> hires within the School of Engineering in the general area of Data
> Science are expected.
>
>
> The position begins Fall 2016. Please browse our website for a
> description of departmental activities and people. To apply go to
>
>
> http://facultysearch.statistics.rice.edu/
>
>
> Applicants are requested to upload a letter of application, curriculum
> vita, graduate transcripts, and reprints/preprints. Please include nam
es
> and addresses, including e-mail, of three individuals who will be
> contacted for letters of recommendation.
>
>
> Inquiries to: phyllis em stat.rice.edu. Include Faculty Search in the
> subject line of the e-mail.
>
>
> The review of applications will begin December 16, 2015. Applications
> will continue to be accepted beyond this date until the position is
> filled. Interviews will begin mid January.
>
>
> Rice University is an Equal Opportunity/Affirmative Action employer an
d
> is committed to increasing the diversity of its faculty.
>
> ------------------------------
> Marina Vannucci
> Professor and Chair
> Department of Statistics
> Rice University
> 6100 Main street
> Houston, TX 77005
> ------------------------------
>
> 9.From: James Frane
> Posted: Tuesday November 3, 2015 5:25 PM
> Subject: RE: Principal Components Analysis
> Message:
> I don't understand how a method of analysis can be recommended without
> understanding precisely what the variables are and what is already kno
wn
> about each individually and what their relationships with each other a
re.
>
>
> There are some things that I don't recommend, e.g., principal componen
ts
> without rotation to simple structure because of a general lack of
> practical interpretation of the unrotated principal components.
> Moreover, by default the rotation should be oblique rather than
> orthogonal, i.e., orthogonal rotation should be used only when oblique
> rotation reveals nearly orthogonal factors. It follows that canonical
> correlation is not recommended because is yields orthogonal
> uninterpretable factors.
>
>
> Dimension reduction may be in order so I accept orthogonally rotated
> prinicpal components (or maximum likelihood factors if you want to go
> elegant) followed by oblique rotation for each of the two sets of
> variables as a possibility, but only after such procedures pass common
> sense acceptance on the basis of the first paragraph above.
> ------------------------------
> James Frane
> Self-Employed
> ----------------------------------------------------------------------
---
> Original Message:
> Sent: 10-30-2015 09:27
> From: Linda Landon
> Subject: Principal Components Analysis
>
> Everyone,
>
> I'm about to make my first foray into principal component analysis sin
ce
> graduate school. I'm working on a multivariate regression problem tha
t
> has multiple independent variables and multiple dependent variables.
> I'm wondering if it would be possible to perform multivariate regressi
on
> by using BOTH the principal components representing dimensions in the
> set of independent variables as independent variables and the principa
l
> components representing dimensions in the set of dependent variables
as
> dependent variables in the same regression model. However, I'm unsure
> if this application is statistically valid.
>
> The common examples found in texts and in the peer-reviewed literature
> when PCA is applied prior to multivariate regression are
>
> (1) multiple independent variables to find the principal components
> in the independent variables and then one or more components are used
as
> independent variables in a regression against a single dependent
> variable or
> (2) to multiple dependent variables to find the principal component
s
> in the dependent variables and then a single component is used as a
> dependent variables in a regression against a more or more independent
> variables.
>
> However, I would like to know if it is possible to perform one
> multivariate regression analysis on two sets of principal components,
> one set of components representing the dimensions in the independent
> variables and the other set of components representing the dimensions
in
> the dependent variables. In my rather limited literature search, I
> haven't found an example of this application in the peer-reviewed
> literature, which could indicate that it is not statistically valid.
> Further, other sources of information don't explicitly address this
> application.
>
> Thanks in advance for your insight.
>
> Linda
>
> Linda A. Landon, PhD, ELS
> President
> Research Communiqu
> Jefferson City, MO
> Email: LandonPhD em ResearchCommunique.com <LandonPhD em ResearchCommunique
.com>
> Phone: 573-797-4517
> Central Standard Time ( CST ) = GMT-6 (November February)
> Central Daylight Time ( CDT ) = GMT-5 (March October)
>
>
>
> You are subscribed to "ASA Connect" as pam em ime.usp.br. To change your
> subscriptions, go to
> http://community.amstat.org/preferences?section=Subscriptions[$AutoLog
in$].
> To unsubscribe from this community discussion, go to
> http://community.amstat.org/HigherLogic/eGroups/Unsubscribe.aspx?UserK
ey=cefeb179-a728-4682-a33a-2a31a03de88b&sKey=48055291112741E690E4&GroupK
ey=4061ce8b-1847-47dd-b1ae-357745003563.
>
>
>
>
> _______________________________________________
> abe mailing list
> abe em lists.ime.usp.br
> https://lists.ime.usp.br/mailman/listinfo/abe
--
*José Ferreira de Carvalho, PhD*
Statistika Consultoria
tel.: +55-19-3236-7537 (Office)
+55-19-98139-9927 (Mobile)
Mais detalhes sobre a lista de discussão abe