在ggplot2中绘制线性回归

Question

I have this dataframe:

> head(data)
    sx   yd  sl
1   male 35  36350
2   male 22  35350
3   male 23  28200
4 female 27  26775
5   male 30  33696
6   male 21  28516

Where "sx" is sex, "yd" is years since acquired degree and "sl" is salary. Using ggplot or plot, I can plot the scatter easily.

palette(c("pink", "blue"))
plot(data$yr, data$sl, col = factor(data$sx), xlab = "Years Since Earned Highest Degree", ylab = "Salary (dollars)", main = "Salary Increases with Experience", pch = 19)
legend("topleft", legend = unique(data$sx), col = c("blue", "pink"), pch=19)

library(ggplot2)
ggplot(data, aes(x=yd,y=sl)) + 
    geom_point(shape=21, aes(col=sx, bg=sx)) + 
    xlab("Years Since Earned Highest Degree") + 
    ylab("Salary (dollars)") + 
    ggtitle("Salary Increases with Experience") + 
    scale_color_discrete(guide=FALSE) + 
    labs(fill="sex")

However, I have also made a linear model based on the data:

mod<-lm(sl~sx*poly(yd,2),data)

And I am unable to figure out how to plot the data to the graphs. Specifically, I want two lines corresponding to male and female data superimposed on the scatterplot AND to be color-coded. I would assume that R has some way to do this so that I don't have to actually write out the model. Either base plot or ggplot answers are good. Thanks.

Edit:

Running this the above ggplot with geom_smooth(aes(col=sx), se = FALSE, method = "lm", formula = sl ~ sx * poly(yd, 2)) :

ggplot(data, aes(x=yd,y=sl)) + geom_point(shape=21, aes(col=sx, bg=sx)) + geom_smooth(aes(col=sx), se = FALSE, method = "lm", formula = sl ~ sx * poly(yd, 2)) + xlab("Years Since Earned Highest Degree") + ylab("Salary (dollars)") + ggtitle("Salary Increases with Experience") + scale_color_discrete(guide=FALSE)+ labs(fill="sex")

Returns this error:

Error in model.frame.default(formula = formula, data = data, weights = weight,  : 
  variable lengths differ (found for '(weights)')
Error in if (nrow(layer_data) == 0) return() : argument is of length zero

Answer 1

data = data.frame(sx = c("male", "male", "male", "female", "male", "male"),
              yr = c(35, 22, 23, 27, 30, 21),
              sl = c(36350, 35350, 28200, 26775, 33696, 28516))
ggplot(data, aes(x=yr,y=sl)) + 
  geom_point(shape=21, aes(col=sx, bg=sx)) + 
  geom_smooth(aes(color = sx), se = FALSE, method = "lm", formula = y ~ poly(x, 2)) + 
  xlab("Years Since Earned Highest Degree") + 
  ylab("Salary (dollars)") + 
  ggtitle("Salary Increases with Experience") +     
  scale_color_discrete(guide=FALSE)+ labs(fill="sex")

Is this what you want? You should get individual fits if you have more data for female. Right now sum(data$sx == 'female') is 1. There's no way to have a polynomial fit to that.
For example, try:

data = data.frame(sx = c("male", "male", "male", "female", "male", "male", "female", "female", "female"),
                  yr = c(35, 22, 23, 27, 30, 21, 25, 18, 29),
                  sl = c(36350, 35350, 28200, 26775, 33696, 28516, 27402, 31492, 23195))

This should work.

Answer 2

I was unable to find a ggplot way to do it, so here is the base plot way to do it:

palette(c("pink", "blue"))
plot(data$yr, data$sl, col = factor(data$sx), xlab = "Years Since Earned Highest Degree", ylab = "Salary (dollars)", main = "Salary Increases with Experience", pch = 19)
legend("topleft", legend = unique(data$sx), col = c("blue", "pink"), pch=19)
lines(seq(0,25,0.1), predict.lm(quad, data.frame(yd = seq(0,25,0.1), sx = "female", stringsAsFactors = TRUE)),col="pink", lwd = 5)
lines(seq(0,25,0.1), predict.lm(quad, data.frame(yd = seq(0,25,0.1), sx = "male", stringsAsFactors = TRUE)),col="blue", lwd = 5)

The two calls to lines are the solution. If anyone has the ggplot way to do it, I'd appreciate it a lot, as ggplot looks so much better.