fake_data.r

library(plyr)
library(ggplot2)

formatter = function(x) formatC(x, format='d', big.mark=',')

# create some fake data
variability = c(A=0.1, B=0.2, C=0.3)

d = as.data.frame(mutate(list(),
  source=sample(c('A', 'B', 'C'), 1000, replace=TRUE),
  x=runif(1000, 200e3, 2e6),
  y=(0.8 + runif(1000, 0, 1) * variability[source]) * x + rnorm(1000, 0, 100e3)
))

# fit a LTO for each source
fits = ddply(d, .(source), function(s) {
  c(slope=coef(lm(y ~ x + 0, data=s))[[1]], intercept=0)
})

# plot it all - multiple lines of fit
p = ggplot(data=d) +
    geom_point(aes(x=x, y=y, colour=source), size=1) +
    geom_abline(data=fits, aes(intercept=intercept, slope=slope, colour=source), size=1) +
    theme_minimal() + labs(x="Market Value", y="Realised Value") +
    scale_x_continuous(labels=formatter) + scale_y_continuous(labels=formatter)
plot(p)


# log-log transformed with log line-hack
slope = coef(lm(y ~ x + 0, data=d))[[1]]
p = ggplot(data=d) +
    geom_point(aes(x=x, y=y)) +
    geom_line(
      aes(x=c(min(x), max(x)), y=c(min(x), slope * max(x))),
      colour='red', size=2
    ) +
    theme_minimal() + labs(x="Market Value", y="Realised Value") +
    scale_x_log10(labels=formatter) + scale_y_log10(labels=formatter)
    #scale_x_continuous(labels=formatter) + scale_y_continuous(labels=formatter)
plot(p)

dev.copy2pdf(file='scatter_example.pdf')