Practical Predictive Analytics

Comparing a sample to the population

To illustrate some of the benefits of sampling, and to see how you can often get close to the same results with a sample as with a larger population, copy the following code and run it within an R script. This script will generate a 15,000,000 row population and then extract a 100-row random sample. Then we will compare the results:

large.df <- data.frame( gender = c(rep(c("Male", "Female", "Female"), each = 5000000)), purchases = c(0:9, 0:5, 0:7) ) #take a small sample y <- large.df[sample(nrow(large.df), 100), ] mean(large.df$purchases) mean(y$purchases) #Render 2 plots side-by-side by setting the plot frame to 1 by 2 par(mfrow=c(1,2)) barplot(table(y$gender)/sum(table(y$gender)))  )barplot(table(large.df$gender)/sum(table(large.df$gender)) ...