Date: Thu, 23 Dec 2004 20:55:26 +0000
Reply-To: iw1junk@COMCAST.NET
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: Ian Whitlock <iw1junk@COMCAST.NET>
Subject: Re: More on the OT programming question
The text below was not posted becasue I left the address to SAS-L in the text.
Does anyone else have this problem? I copy text from SAS-L into a Comcast message and add my message in front.
David,
Back in the 1950's I attended a talk by von Braun about going
into space. At the end he took question from the audience.
One was, "Are there any unknown dangers in space?" The answer
was, "Yes."
Peter,
Was it the government that didn't give you a fundable score, or
your employer? Would your proposed models conflict with current
administration "science"?
I would concentrate on what you think is wrong with current
models and the consequences of working with them, preferably
quantified in terms of money otherwise human lives.
One common approach to the unknown is to promise research already
done and spend the grant money on the next grant proposal, i.e. if
your situation permits do the work after hours to get a base for
the future.
Ian_Whitlock@comcast.net
=============================
Date: Wed, 22 Dec 2004 13:28:22 -0800
Reply-To: cassell.david@EPAMAIL.EPA.GOV
Sender: "SAS(r) Discussion"
From: "David L. Cassell" <cassell.david@EPAMAIL.EPA.GOV>
Subject: Re: More on the OT programming question
In-Reply-To: <s1c92dcb.068@MAIL.NDRI.ORG>
Content-type: text/plain; charset=US-ASCII
Peter Flom <flom@NDRI.ORG> replied:
> Here's the detail
>
> Where I work, we do grant-funded research (usually funded by NIH or
> NIDA - i.e. US federal agencies). I made a proposal to work on models
> for one-inflated negative binomial and one-inflated Poisson
regression.
> They generally liked the proposal, but it didn't get a fundable score.
> So I'd like to resubmit. One of the critiques was that the models I
> propose to develop are similar to existing models, and that "the
> application does not make clear where additional statistical and
> programming problems arise and how they will be solved"; in addition,
> they wanted better justification for the amount of time I wanted in my
> budget.
>
> How can I know what problems arise before they arise? And, since I
> don't know what problems will arise, how can I know in advance how
they
> will be solved? And, given that I don't know in advance what problems
> will arise, how can I justify the amount of time it will take to solve
> them? I mean, at least some of the problems will require insight.
How
> can I predict time to insight? I might see the solution right away,
or
> only after lots of time.....
>
> Have others dealt with this kind of thing, either from funding
agencies
> or from bosses?
Is there an 'accepted' approach for statistical research proposals
there?
Are there prior (accepted) proposals in the same line that you could
study
in order to see what is considered a reasonable answer to the above
questions?
If someone else has done this and gotten funding, their proposal may
have
the right buzzwords and such that the reviewers are looking for. You
know
where the shortfalls are in the current models. Play those up. You
know
what directions you want to work in to improve the current models. Lay
those
directions out step-by-step if possible, with as much forethought as
possible
into what could go wrong. Those are as close to the unknown problems as
you
can guess right now.
Unfortunately, there may be no right answer, depending on the reviewers
and the judging criteria. In a former life, I had a similar problem. I
never did get the funding, since there was no way I could (or would)
guarantee
that I would have a solution to the stated stat problem in a fixed
amount of
time. My Pointy-Haired Boss couldn't grasp that I couldn't predict
ahead of
time what might be unpredictable. :-(
HTH,
David
--
David Cassell, CSC
Cassell.David@epa.gov
Senior computing specialist
mathematical statistician
| 
