R help - Mar 2007 - Sigmoidal fitting

I am trying to write a function that fits a sigmoid given a X and Y vector
guessing the start parameters.
I use nls. What I did (enclosed) seems to work well with many data points but if
I want to fit small
vectors like :

pressure <- c(5,15,9,35,45)
gas <- c(1000,2000,3000,4000,5000)

it do not work. The help page says that it do no not work on zero residual data.

Massimo Cressoni

On Sat, 2007-03-03 at 16:30 -0500, Cressoni, Massimo (NIH/NHLBI) [F]
wrote:
> I am trying to write a function that fits a sigmoid given a X and Y vector
guessing the start parameters.
> I use nls. What I did (enclosed) seems to work well with many data points
but if I want to fit small
> vectors like :
> 
> pressure <- c(5,15,9,35,45)
> gas <- c(1000,2000,3000,4000,5000)
> 
> it do not work. The help page says that it do no not work on zero residual
data.
> 
> Massimo Cressoni
Hi Massimo

I think which this script solve your question
------------------------
pressure <- c(5,9,15,35,45)
gas <- c(1000,2000,3000,4000,5000)
plot(gas,pressure)
model.1 <- nls(pressure ~ SSlogis(gas, ASym, xmid, scal)))
coef.sig<-coef(summary(model.1))[,1]
est.p<-coef.sig[1]/(1+exp((coef.sig[2]-gas)/coef.sig[3]))
points(gas,est.p,col=2)
------------------------

If you need more help just need a mail


-- 
Bernardo Rangel Tura,M.D.,Ph.D
National Institute of Cardiology
Rio de Janeiro - Brazil

I want to thank professor Tura for his help but the function in some cases fails
to converge and I am not able to
understand the reason :
> GAS
[1] 0.8108649 1.0386906 1.1638837 2.4144286
> PRESSURE
[1]  0  5 15 45
>  model.1 <- nls(GAS_PER_G ~ SSlogis(PRESSURE, ASym, xmid, scal))
 
 model.1 <- nls(GAS_PER_G ~ SSlogis(PRESSURE, ASym, xmid, scal))
Error in nls(y ~ 1/(1 + exp((xmid - x)/scal)), data = xy, start = list(xmid =
aux[1],  :
	step factor 0.000488281 reduced below 'minFactor' of
0.000976562
> 
Massimo Cressoni