knitr::opts_chunk$set( collapse = TRUE, comment = "#>" )
The explore package offers an easy way to do basic A/B testing.
For interactive A/B testing simply call abtest()
without any parameters.
{width=600px}
If you want to A/B test your own data, pass them into the ´abtest()´ function.
In this example we are using synthetic data using one of the create_data_()
functions of {explore}
library(dplyr) library(explore) data <- create_data_buy(obs = 1000)
data %>% describe()
Each observation is a customer. The ´buy´ variable contains 0/1 values (1 = the customer did buy a product)
We want to test the hypothesis, that customer with age > 50 have a different buying behavior.
data %>% explore(age, target = buy)
We see a clear pattern, that people who buy differ in age from people who don´t buy. But is this difference statistically significant?
data %>% abtest(age > 50, target = buy, sign_level = 0.05)
The A/B test shows a statistically significant difference!
We would like to test the hypothesis, that customer with a mobile voice product (mobilevoice_ind == 1) have different bbi_usg_gb (broadband usage in GB)
data %>% explore(mobilevoice_ind, target = bbi_usg_gb)
The boxplot shows a small difference in Broadband Usage between customers with/without Mobile Voice product. But is this difference statistically significant?
data %>% abtest(mobilevoice_ind == 1, target = bbi_usg_gb, sign_level = 0.05)
The A/B test shows a NOT statistically significant difference! The p-value is 0.09, but should be max. 0.05 (as we defined sign_level = 0.05)
In this example we are using the Titanic datasst. use_data_titanic()
is a functions of {explore}
that makes it easier to use popular example datasets.
library(dplyr) library(explore) data <- use_data_titanic(count = TRUE)
data %>% describe()
We want to test the hypothesis, that female passengers have a higher chance to survive.
data %>% explore(Sex, target = Survived, n = n)
We see a clear pattern, but is it difference statistically significant?
data %>% abtest(Sex == "Female", target = Survived, n = n, sign_level = 0.05)
The A/B test shows a statistically significant difference!
Any scripts or data that you put into this service are public.
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.