Search code examples
rggplot2curve-fittingdata-fittingnls

Fitting a nls with ggplot2 - error object of type 'symbol' is not subsettable

I am trying to do a nls fit with ggplot with the following data: df <- data.frame(t = 0:30, m = c(125.000000, 100.248858, 70.000000, 83.470795, 100.000000, 65.907870, 66.533715, 53.588087, 61.332351, 43.927435, 40.295448, 44.713459, 32.533143, 36.640336, 40.154711, 23.080295, 39.867928, 22.849786, 35.014645, 17.977267, 21.159180, 27.998273, 21.885735, 14.273962, 13.665969, 11.816435, 25.189016, 8.195644, 17.191337, 24.283354, 17.722776)

The code I have so far (a bit simplified) is 

ggplot(df, aes(x = t, y = m)) +
geom_point() +
geom_smooth(method = "nls", formula=log(y)~x)

But I get the following error

Error in cll[[1L]] : object of type 'symbol' is not subsettable

I have browsed stackoverflow and found similar issues, but I have not been able to solve the problem. I would really like to plot the data without transforming the axis.

Any help is much appreciated.

Solution

  • For nls you have to specify the parameters more carefully. Also, you probably don't want to use log(y), because that will plot the logarithm instead of y. My guess is that you want to use something like y ~ exp(a + b * x). See below for an example. 

    ggplot(df, aes(x = t, y = m))+
  geom_point()+
  geom_smooth(method = "nls", formula = y  ~ exp(a + b * x), start=list(a=0, b=1), se=FALSE)

    You could also use glm instead of nls. For instance the following gives you the same solution. 

    ggplot(df, aes(x = t, y = m))+
  geom_point()+
  geom_smooth(method = "glm", formula = y  ~ x, family=gaussian(link = "log"), se=FALSE)