[R] Fitting function with if-clause (nls; e.g. heaviside)

[Prof Brian Ripley]

On Wed, 23 Jun 2004, Karl Knoblick wrote:

> I want to fit a function. The function is e.g.:
> y = c+m1*x if x<0, c+m2*x if x>=0
> where m1, m2 and c is a parameter and x, y are
> variables of a data frame.
> 
> I think using  nls is appropriate. But I do not know,
> how to type this formula in nls. Can anybody help?

It's a linear model (linear in the params)

lm( y ~ I(x*(x>0) + I(x*(x<0)))

although I would define some new variables to make that easier to read, 
e.g.

xplus <- x*(x>0)
xminus <- x*(x<0)
lm(y ~ xplus + xminus)

-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595