# Can I Use a Dummy Variable in Quadratic form in lm() in R?

I'm trying to replicate models published in The Age-Productivity Gradient: Evidence from a Sample of F1 Drivers and am stumbling at first hurdle over a line "the baseline specification contains a quadratic in age and drivers dummies".

I can get quadratic age (a numerical var) and linear name dummy (a factor in the original dataset) with the following construction:

lm(pos ~ name + I(age) +I((age)^2),data=ergastData)


but I get an error on

lm(pos ~ name + I(name^2) + I(age) +I((age)^2),data=ergastData)


because name is a factor and squaring it is not meaningful?

So how can I get a quadratic term for the dummy variable?

(I'm not a statistician.. I assume this is a meaningful thing to do and my interpretation of the original paper is correct!)

As an extension of this question, with ggplot2, I can chart the individual name model results using:

g = ggplot(egastdata, aes(x=age,y=pos,col=name))
g = g + stat_smooth(method = "lm", formula = y ~ I(x) +I((x)^2)) +geom_point()
g


This seems to group on name and presumably generate separate models for each driver (i.e. a set of distinct age coefficients for each driver). How would I use ggplot to chart the result of a single set of age coefficients and then separate name dummy variable coefficients (both linear in name and quadratic)? (That is, I want a single model around age with a separate coefficients around name for each driver.) Would the age coefficients be the same as if I ran the model over the whole set of driver data without distinguishing drivers?

• Read the "and" like "plus". They include a variable $age^2$ plus a variable $driver dummy$. That is, "the baseline specification contains (a quadratic in age) + (drivers dummies)". – coffeinjunky Apr 26 '14 at 13:02