[Prévia] [Próxima] [Prévia por assunto] [Próxima por assunto]
[Índice cronológico]
[Índice de assunto]
[Fwd: Summary - How to look after your statistician]
- Subject: [Fwd: Summary - How to look after your statistician]
- From: Frederico Zanqueta Poleto <fred@poleto.com>
- Date: Mon, 22 Mar 2004 10:51:05 -0300
A mensagem que estou encaminhando cont�m resultados de uma pesquisa espont�nea
feita entre os membros da lista de discuss�o "allstat" sobre problemas que
costumam acontecer em assessorias estat�sticas.
Pelo que vejo em conversas com profissionais brasileiros, elas n�o s�o muito
diferentes do que acontece por aqui. Uma leitura interessante.
--
Frederico Zanqueta Poleto
fred@poleto.com
--
"An approximate answer to the right problem is worth a good deal more than an exact answer to an approximate problem." J. W. Tukey
-------- Original Message --------
Many Thanks to all of you who sent their pet gripes. Several sent
informative/amusing/depressing stories, which unfortunately, to preserve
anonymity, can't easily be reproduced!
Several of you have also asked for copies of the talk that I am giving.
I'm afraid that as I've never really mastered powerpoint and tend to
largely 'make up lectures' as I go, I don't really have anything that I
can send out. Hopefully the summary below will make up for that.
A point made in several replies was that statisticians need to make an
effort to get on with clients, in particular to try and speak a language
that the client will understand. I'm sure that we'd all agree that
problems in statistical consultancy aren't all one sided - they just
feel like it some times!
I've summarized below as well as I could. Perhaps some responses should
fall into different 'categories' but hopefully I've got them broadly
correct. The numbers in parenthesis indicates how many responses of
each type I received (most people sent several responses). I've phrased
the responses 'negatively' "we are unhappy with" .... many replies were
in fact positively phrased "it would be better if clients ..." The
negative slant a) feels more natural to me somehow (wonder what that
says about my state of mind!) b)probably reflects more generally the
'tone' of the responses I received. If somehow I've missed out your
contribution - my apologies, it wasn't intentional.
Once again, thanks for you help with this.
Graham
=======================================================================
We are unhappy with �
(20) Not being consulted from the start, when we could help develop a
well designed study and subsequently not have to sort out the mess after
the event.
(12) Clients who take the statistician �for granted�. For instance
never saying thanks
(10) Not being credited as an author on a paper, or a collaborator or
co-applicant in grant despite undertaking significant work
(10) People who are going to ignore our advice and use whatever
statistical approach they like � sometimes because �Blogs uses that
method�; sometimes based upon a very limited short stats course they
took as an undergraduate
(7) People who ask for sample size calculations but don't provide any of
the necessary information to undertake this.
(5) People who fail to realize or accept a non-significant result
(either that it is �statistically� non-significant, or �functionally�
unimportant).
(5) People who don't give the background/explanation of the context in
which the data was collected
(5) Clients who give very short (sometimes false) deadlines.
(4) Being asked to sort out the software/the data base � computational
things which should have been sorted out by someone else earlier.
(4) Clients who don't have any question/objective in mind when they hand
you the data.
(3) People who misrepresent/change the results/graphs etc. after you
have done the analysis, or put pressure on you to �change� the results
(3) People using tests they don't understand and/or programs whose
printout they can't interpret.
(18) Odds & Ends (only got a few mentions)
We are unhappy with clients who �
Go to other statisticians for advice once we have started working with them.
Fail to appreciate that the statistician may know a fair amount about
the topic area
Fail to appreciate that statisticians may specialize in their knowledge
and failing to discriminate between a RA and a Professor's expertise.
Who ask us about a really �unusual� test then look at us �weirdly� when
we haven't heard of it
Who collect �millions� of pieces of data on a limited number of subjects
Who try to �bribe� us (with flowers/drink)
Who only want a P value
Who don't appreciate that correlation isn't causation
Or that 100:300 is not �the same as� 1:3
Or that once the subject is randomized they are �in the study�
Or don't realize that their �simple question� isn't!
And many, many of use share the �heart sink� feeling when the client says;
�This will only take 2 minutes�.
--
Dr G.S.Clarke
Lecturer in Physiology & Biometery
Faculty of Health Studies
University of Wales, Bangor
Fron Heulog
Ffriddoedd Road
Bangor
Gwynedd LL57 2EF
Tel: 01248 383157
e-mail: g.clarke@bangor.ac.uk