Cubic Regression in R

Question

I'm trying to fit a cubic curve to my scatterplot. I was able to do this in minitab with no problem, but I'm finding it quite difficult to fit a cubic nonlinear regression to my data. my data is set up like:

score 3 · Accepted Answer · answered Aug 27 '13 at 18:59

3

See ?poly which (by default) sets up a design matrix-ready orthogonal polynomial of the specified degree over the range of x-values you hand to it:

df <- read.table(text="AGE Value
3   10
4   10
5   11
5   13
6   10
7   9
8   8", header=TRUE)

lm(Value ~ poly(AGE, degree=3), data=df)

# Call:
# lm(formula = Value ~ poly(AGE, 3), data = df)
# 
# Coefficients:
#   (Intercept)  poly(AGE, 3)1  poly(AGE, 3)2  poly(AGE, 3)3  
#       10.1429        -2.0026        -2.3908         0.6019

answered Aug 27 '13 at 18:59

Josh O'Brien

159,210
26
366
455

Or use `polyfit` tho' granted that does not return information about the fit quality. – Carl Witthoft Aug 27 '13 at 19:38
Or, if you're after a (B-)spline fit, you could use `bs(AGE, degree=3)` (from the **splines** package) in your formula, setting the total `df` [as demo'd here](http://stackoverflow.com/questions/12925533/quadratic-spline/12938310#12938310). – Josh O'Brien Aug 27 '13 at 19:40
@CarlWitthoft -- Where would I find `polyfit`? – Josh O'Brien Aug 27 '13 at 19:41
Well, that was silly of me. I apologize: `pracma::polyfit` . – Carl Witthoft Aug 27 '13 at 21:06

Cubic Regression in R

1 Answers1

Linked