How to deal with a quadratic model that has too many fitted values?

Question

I'm trying to fit a quadratic regression model to a dataset and then plot the curve on a scatterplot. The dataset is about number of episodes and screentime for characters in a TV show.

I plotted a scatterplot with episodes on x axis and screentime on y axis this worked fine.

Then I create the model as follows:

#ordering
gottemp <- got[order(got$episodes),]

#plotting
plot(screentime~episodes, data = gottemp, xlab ="Number of episodes", ylab = "Screentime (minutes)", col=c("blue","red")[gender], pch=c(1,2)[gender])
legend("topleft",pch = c(1,2),col=c("blue","red"),c("female","male"))
title("Plot of Screentimes vs Number of Episodes")

#creating 3model and plotting line
model <- lm(screentime~episodes+I(episodes^2), data = got)
lines(fitted(model))

This gives me a model with correct coeefficients however the line that is plotted is not what would be expected. When I view the model i see that there are 113 fitted values, which I think is due to some characters having the same number of episodes so to fix this I think there should only be one fitted value for each number of episodes.

Answer 1

Something like

nd <- data.frame(episodes=seq(min(episodes), max(episodes), length=51)
nd$screentime <- predict(model, newdata=nd)
with(nd, lines(episodes, screentime))

should do what you want. There's probably a duplicate around somewhere...