nls2 is an R package that adds the "brute-force" algorithm and multiple starting values to the R nls function. nls2 is free software licensed under the GPL and available from CRAN. It provides a function, nls2, which is a superset of the R nls function which it, in turn, calls.

News

August 22, 2010. nls2 version 0.1-3 uploaded to CRAN. See NEWS file.

Installation

As with any R package on CRAN, nls2 can be installed from within R using

install.packages("nls2")

Help

After installation, the nls2 help page can be accessed from within R via

library(nls2)
help(nls2)

or it can be accessed online here.

Examples

There are examples:

• at the end of the nls2 help page.
• posted on r-help: see this question and its answer.
• on pages 28 and 29 of the book Nonlinear Regression with R by Ritz & Streibig.

Citation

The citation for this package can be obtained by issuing this command from within R:

citation("nls2")

FAQs

Q. Why does this not work?

library(nls2)

# data is from ?nls
DNase1 <- subset(DNase, Run == 1)
fm1DNase1 <- nls(density ~ SSlogis(log(conc), Asym, xmid, scal), DNase1)

# next line is wrong
predict(as.lm(fm1DNase1), newdata = list(conc = 1:12))

A. The problem is that the variables in the tangent linear model are Asym, xmid and scal -- not conc. One can only predict at the fitted points although one can interpolate among those -- in the next version of nls2 as.lm will have a newdata= argument which will generate a new lm based on a randomly synthesized response so that predict(as.lm(fm1DNase1, newdata = list(conc = 1:12), interval = "confidence") will be possible but in the meantime:

library(nls2)

# run nls
DNase1 <- subset(DNase, Run == 1)
fm1DNase1 <- nls(density ~ SSlogis(log(conc), Asym, xmid, scal), DNase1)

# predict at original points
pred <- predict(as.lm(fm1DNase1), interval = "prediction")

# interpolate
x0 <- 1:12
pred0 <- apply(pred, 2, function(y) spline(DNase1$conc, y, xout = x0)$y)

# plot
plot(density ~ conc, DNase1)
matlines(DNase1$conc, pred)
matpoints(x0, pred0, col = 1:3, pch = 20)

(Note that this package is unrelated to the software of the same name associated with the book Statistical Tools for Nonlinear Regression by Huet et al.)