Nonlinear Least Squares in S-plus and R

Steve Simon

2004-06-28

Categories: Blog post Tags: Nonlinear regression

I've worked on a brief explanation of how to fit an S-shaped curve using SPSS and someone wanted to know how to do this in S-plus. Here's a simple example.

xc <- c(0.8,0.5,0.2,-0.2,-0.5,-0.8) yn <- c(0.02,0.10,0.29,0.20,0.80,1.00) nls.mod1 <- nls(yn~1/(1+exp(b*xc-a)),start=list(a=0,b=10)) summary(nls.mod1) Formula: yn ~ 1/(1 + exp(b * xc - a)) Parameters: Value Std. Error t value a -2.70380 1.4019 -1.92867 b 7.90714 3.7126 2.12981 Residual standard error: 0.149848 on 4 degrees of freedom Correlation of Parameter Estimates: a b -0.908 plot(xc,yn,xlim=c(-1,1)) x.grid <- seq(-1,1,by=0.01) lines(x.grid,predict(nls.mod1,newdata=list(xc=x.grid)))

This same code should also work in R.

You can find an earlier version of this page on my original website.