crime_report_vs_temperature.R

# Read in the data
days <- read.csv('all_years.csv')

# Plot all the data
plot(total_count ~ temp_max, days, col=rgb(0,0,0, .3))

# Fit a cubic polynomial to the data
m1 <- lm(total_count ~ temp_max + I(temp_max^2) + I(temp_max^3), days)

# Extract the coefficents
b_0 <- coef(m1)[1]
b_1 <- coef(m1)[2]
b_2 <- coef(m1)[3]
b_3 <- coef(m1)[4]

# Plot the curve
curve(b_0 + b_1*x + b_2*x^2 + b_3*x^3,
      min(days$temp_max),
      max(days$temp_max),
      add=TRUE,
      col="red")


# Output as a png
pdf("crime_vs_temp.png")
plot(total_count ~ temp_max, days, col=rgb(0,0,0, .3))
curve(b_0 + b_1*x + b_2*x^2 + b_3*x^3,
      min(days$temp_max),
      max(days$temp_max),
      add=TRUE,
      col="red")
dev.off()

The decline in crime at higher temperatures is much smaller.

Also notice that instead of one cloud of points it looks like there's two. This calls out for more investigation.