exponential fit in ggplot R

Question

I've been trying to fit an exponential curve to my data using ggplot and geom_smooth. I'm trying to replicate the answer to a similar problem (geom_smooth and exponential fits) but keep getting following error message:

> exp.model <-lm(y ~ exp(x), df)
Error in lm.fit(x, y, offset = offset, singular.ok = singular.ok, ...) : 
  NA/NaN/Inf in 'x'

I don't understand the error, as there is not NA/NaN/Inf values in the dataset:

>df
      x         y
1  1981  3.262897
2  1990  2.570096
3  2000  7.098903
4  2001  5.428424
5  2002  6.056302
6  2003  5.593942
7  2004 10.869635
8  2005 12.425793
9  2006  5.601889
10 2007  6.498187
11 2008  6.967503
12 2009  5.358961
13 2010  3.519295
14 2011  7.137202
15 2012 19.121631
16 2013  6.479928

Koph

I've seen that post and tried to replicate the plot, but got the error message again.. – Koph

@nrussell you marked this post as duplicate, but I get an error message trying to replicate the answer in the referred post. I've clarified above. – Koph

Exponentiating such large numbers will result in floating-point overflows. hint: try exp.model <-lm(y ~ exp(x-1981), df) – Bethsaida

Answer 1

Set up data:

dd <- data.frame(x=c(1981,1990,2000:2013),
  y = c(3.262897,2.570096,7.098903,5.428424,6.056302,5.593942,
  10.869635,12.425793,5.601889,6.498187,6.967503,5.358961,3.519295,
  7.137202,19.121631,6.479928))

The problem is that exponentiating any number larger than about 709 gives a number greater than the maximum value storable as a double-precision floating-point value (approx. 1e308), and hence leads to a numeric overflow. You can easily remedy this by shifting your x variable:

lm(y~exp(x),data=dd) ## error
lm(y~exp(x-1981),data=dd) ## fine

However, you can plot the fitted value for this model more easily as follows:

library(ggplot2); theme_set(theme_bw())
ggplot(dd,aes(x,y))+geom_point()+
   geom_smooth(method="glm",
            method.args=list(family=gaussian(link="log")))
            
            

Bethsaida

Thanks, this is pretty much what I was looking for! – Koph

here's another example that may be of use here – Freeloader