R help - Mar 2004 - Terminology and canonical statistical user literature

Brian Ripley wrote (to somebody asking about "effect
sizes"):
> ...
> Given that, I wonder if you are used to standard terminology.
Good point. But I think for many of us there is more behind that.

I personally belong to an (apparently fairly large) group of
R users who may be enthusiastic, but are statistical laymen
due to a lack of formal education in the area.

The half-knowledge that I have is often sufficient to see that 
many otherwise nice sources of statistical knowledge are 
dangerously incomplete when it comes to explaining the
preconditions required for applying a certain technique
(One example: The extensive NIST handbook at
 http://www.itl.nist.gov/div898/handbook/
 fails to mention that the Wilcoxon rank sum test assumes a
 continuous distribution underlying the sample)
This is not to speak of how to correctly interpret the results.

My situation is this:
- I often have a hard time understanding the R documentation 
due to lack of background.
- I am not in a position to obtain a full background like 
a statistics student would get it.
- I am very interested in carefully checking/validating my 
application of statistical techniques.
- I cannot usually get a consulting statistician to help me.

My question:
Could some of the R gurus maybe agree on a book 
(or very small set of books) with the following properties?:
- explains typical approaches of statistical analysis
  (like MASS, but not as condensed)
- carefully describes preconditions, how to check them,
  robustness if they are violated, interpretation of results
- avoids explaining the innards of the techniques
  (and generally uses the perspective of the computer age)
- uses terminology that is easily mapped to R

If yes, I would be very interested in seeing this list.

I understand that one book cannot cover it all,
but maybe there is at least something like "CAS-"
(Conservative Applied Statistics without S) that
is of this type?  :-)

  Lutz Prechelt

Prof. Dr. Lutz Prechelt;  prechelt at inf.fu-berlin.de
Institut f?r Informatik; Freie Universit?t Berlin
Takustr. 9; 14195 Berlin; Germany
+49 30 838 75115; http://www.inf.fu-berlin.de/inst/ag-se/

Dr. Prechelt  -

It's been my observation that there IS no book of the sort
you have asked for.  There have been many attempts over the
last 75 years to write such a book.  Attempts by some very
smart and articulate people  . . .  and no such attempt that
I know of has succeeded.

I am forced to conclude that there is something intrinsic to
the subject matter which makes it refractory to a good textbook
exposition.  We can speculate about what that is, but I think
the evidence is plain that there is some inherent difficulty.

The way which does seem to work in learning this material is
a spiral - do an introductory look with a limited set of basic
statistical procedures.  Don't worry if you didn't understand
quite all of the details.  Let it sit for a few months;  find
an applied situation where you just HAVE to make use of what
you know.  Then, three months later, go back and view all of
the same procedures again, from a somewhat more sophisticated
or abstract viewpoint, and extend your knowledge to a few more
procedures.

This approach to learning statistics does seem to work, but
it's not a quick process.  I don't know of any other.

-  tom blackwell  -  u michigan medical school  -  ann arbor  -

On Tue, 16 Mar 2004, Lutz Prechelt wrote:
> Brian Ripley wrote (to somebody asking about "effect sizes"):
> > ...
> > Given that, I wonder if you are used to standard terminology.
>
> Good point. But I think for many of us there is more behind that.
>
> I personally belong to an (apparently fairly large) group of
> R users who may be enthusiastic, but are statistical laymen
> due to a lack of formal education in the area.
>
> The half-knowledge that I have is often sufficient to see that
> many otherwise nice sources of statistical knowledge are
> dangerously incomplete when it comes to explaining the
> preconditions required for applying a certain technique
> (One example: The extensive NIST handbook at
>  http://www.itl.nist.gov/div898/handbook/
>  fails to mention that the Wilcoxon rank sum test assumes a
>  continuous distribution underlying the sample)
> This is not to speak of how to correctly interpret the results.
>
> My situation is this:
> - I often have a hard time understanding the R documentation
> due to lack of background.
> - I am not in a position to obtain a full background like
> a statistics student would get it.
> - I am very interested in carefully checking/validating my
> application of statistical techniques.
> - I cannot usually get a consulting statistician to help me.
>
> My question:
> Could some of the R gurus maybe agree on a book
> (or very small set of books) with the following properties?:
> - explains typical approaches of statistical analysis
>   (like MASS, but not as condensed)
> - carefully describes preconditions, how to check them,
>   robustness if they are violated, interpretation of results
> - avoids explaining the innards of the techniques
>   (and generally uses the perspective of the computer age)
> - uses terminology that is easily mapped to R
>
> If yes, I would be very interested in seeing this list.
>
> I understand that one book cannot cover it all,
> but maybe there is at least something like "CAS-"
> (Conservative Applied Statistics without S) that
> is of this type?  :-)
>
>   Lutz Prechelt
>
> Prof. Dr. Lutz Prechelt;  prechelt at inf.fu-berlin.de
> Institut f?r Informatik; Freie Universit?t Berlin
> Takustr. 9; 14195 Berlin; Germany
> +49 30 838 75115; http://www.inf.fu-berlin.de/inst/ag-se/
>
> ______________________________________________
> R-help at stat.math.ethz.ch mailing list
> https://www.stat.math.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide!
http://www.R-project.org/posting-guide.html
>

I would recommend Peter Dalgaard's book for an introduction to Statistics 
with R. Also, a resource that maybe more aligned with what you are asking 
for may be the so called ARTIST project. Details at
www.gen.umn.edu/artist/






"Lutz Prechelt" <prechelt at pcpool.mi.fu-berlin.de>
Sent by: r-help-bounces at stat.math.ethz.ch
03/16/2004 06:44 AM

 
        To:     "R Help" <r-help at stat.math.ethz.ch>
        cc: 
        Subject:        [R] Terminology and canonical statistical user
literature


Brian Ripley wrote (to somebody asking about "effect
sizes"):
> ...
> Given that, I wonder if you are used to standard terminology.
Good point. But I think for many of us there is more behind that.

I personally belong to an (apparently fairly large) group of
R users who may be enthusiastic, but are statistical laymen
due to a lack of formal education in the area.

The half-knowledge that I have is often sufficient to see that
many otherwise nice sources of statistical knowledge are
dangerously incomplete when it comes to explaining the
preconditions required for applying a certain technique
(One example: The extensive NIST handbook at
http://www.itl.nist.gov/div898/handbook/
fails to mention that the Wilcoxon rank sum test assumes a
continuous distribution underlying the sample)
This is not to speak of how to correctly interpret the results.

My situation is this:
- I often have a hard time understanding the R documentation
due to lack of background.
- I am not in a position to obtain a full background like
a statistics student would get it.
- I am very interested in carefully checking/validating my
application of statistical techniques.
- I cannot usually get a consulting statistician to help me.

My question:
Could some of the R gurus maybe agree on a book
(or very small set of books) with the following properties?:
- explains typical approaches of statistical analysis
(like MASS, but not as condensed)
- carefully describes preconditions, how to check them,
robustness if they are violated, interpretation of results
- avoids explaining the innards of the techniques
(and generally uses the perspective of the computer age)
- uses terminology that is easily mapped to R

If yes, I would be very interested in seeing this list.

I understand that one book cannot cover it all,
but maybe there is at least something like "CAS-"
(Conservative Applied Statistics without S) that
is of this type?  :-)

Lutz Prechelt

Prof. Dr. Lutz Prechelt;  prechelt at inf.fu-berlin.de
Institut f?r Informatik; Freie Universit?t Berlin
Takustr. 9; 14195 Berlin; Germany
+49 30 838 75115; http://www.inf.fu-berlin.de/inst/ag-se/

______________________________________________
R-help at stat.math.ethz.ch mailing list
https://www.stat.math.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html

--
This message has been scanned for viruses and
dangerous content by MailScanner, and is
believed to be clean.

On Tue, 16 Mar 2004, Lutz Prechelt wrote:
> Brian Ripley wrote (to somebody asking about "effect sizes"):
> > ...
> > Given that, I wonder if you are used to standard terminology.
> 
> Good point. But I think for many of us there is more behind that.
But you have completely missed my point. Asking for how to do X, where X 
is a word or two, without any reference or explanation as to what X is, 
will be insufficient unless completely standard jargon is used.  Jargon is 
great for concise and accurate transmission of ideas, but only if both 
sides have the same dictionary.

This is particularly important with a one-way transmission medium like 
posting to a list.

It is like the Q yesterday about the `norm of a complex number'.  That's
not standard, but saying it was |x| told us he meant the modulus.

-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595