Path: news.mathworks.com!not-for-mail
From: "John D'Errico" <woodchips@rochester.rr.com>
Newsgroups: comp.soft-sys.matlab
Subject: Re: need exponential growth function with features
Date: Fri, 6 Nov 2009 00:04:01 +0000 (UTC)
Organization: John D'Errico (1-3LEW5R)
Lines: 30
Message-ID: <hcvp5h$ato$1@fred.mathworks.com>
References: <hcvkl5$82f$1@fred.mathworks.com>
Reply-To: "John D'Errico" <woodchips@rochester.rr.com>
NNTP-Posting-Host: webapp-02-blr.mathworks.com
Content-Type: text/plain; charset="ISO-8859-1"
Content-Transfer-Encoding: 8bit
X-Trace: fred.mathworks.com 1257465841 11192 172.30.248.37 (6 Nov 2009 00:04:01 GMT)
X-Complaints-To: news@mathworks.com
NNTP-Posting-Date: Fri, 6 Nov 2009 00:04:01 +0000 (UTC)
X-Newsreader: MATLAB Central Newsreader 869215
Xref: news.mathworks.com comp.soft-sys.matlab:582887


"Matthew" <matthew.merritt.nospam@utsouthwestern.edu> wrote in message <hcvkl5$82f$1@fred.mathworks.com>...
> Hi,
> 
> I am trying to model the bolus delivery of a pharmeceutical. Theoretically, the delivery shold look like:
> 
> 1-exp(-rate*abscissa);
> 
> However, we are actually delivering the bolus through a syringe with the plunger being pushed by a human, i.e., it is hardly constant. When I try to model the data, I get back a rate, but the fit doesn't seem to model the data very well. It tends to peak a little early, then last longer than it should. 
> 
> I suspect that I need a model for the injection that has more detail associated with it. Is their a function that would produce an exponential growth with small oscillations in it that could be controlled by very few variables? I am trying to use fminsearch, and I don't want to have to search over a large number of parameters to get a decent fit, but I suspect only fitting the rate is insufficient. 
> 
> Any ideas?

No, here are no SIMPLE functions that can handle
this class of model. So why not use something
a bit more generic? You can still compute an
effective (time varying) rate constant.

Use my SLM tools to build a spline model, that is a
monotone decreasing function. You can even force
it to go through the point (x,y) = (0,1) if you wish.

Once you have the function fit as a spline, you can
then use slmeval to evaluate both the spline and
its derivative at many points along the curve. Plot
the function f'(x)/(1-f(x)). This would be a local
effective rate. 

HTH,
John