R Tutorial: A/B Testing in R

Key Takeaways

hi my name is Paige peach Nene I'm a data scientist and I'll be your instructor for this course on AV testing in our a/b testing is a powerful way to try out a new design or program changes before making file decisions in this course we'll go over the fundamentals of a/b testing so you can get started on building and analyzing your own AV experiments before getting into a/b testing let's talk about what it is and why it's useful for you a B testing is a framework for you to test different ideas for how to improve upon an existing design often a website with a/b testing you're able to take a set of new ideas test them with a new experiment cicely analyze the results to confidently say which idea is better update your website or app to use the winning idea and then continue the cycle over again what's key to remember is a B testing is not something you do just once you want to be constantly updating your website or app to maximize things like conversion rates or usage time with a be testing you will always be making minor updates to push those numbers up while AV testing is often discussed in the context of websites and tech really it can be used in any context free of a question you want to test and then make updates accordingly a B testing is just experimental design you could a be test two different fertilizer types in your garden or secretly test two coffee brands at work to see which people liked more the world is your a be testing playground in Chapter three I'll go over more example uses a baby testing now let's walk through a simplified set of steps with a hypothetical experiment in future chapters we'll see how a B testing can be more complicated than our hypothetical example here we'll be covering a/b testing concepts in more depth in chapters 3 & 4 let's say you run a cat adoption website right now your home page looks like this you want to know if a different homepage picture would result in more visitors clicking the adopted a button this is also referred to as conversion rate if someone clicks you say they converted the conversion rate is generally the number of people who did an action for example click the button divided by the number of people who went to the page in our case to test this you need two conditions one a control your current photo and two a test a new photo for your test photo you decide to use this photo your hypothesis is that seeing a cat in a hat will make people more likely to want to adopt let's go over the variables that you know you have your question while changing the home page photo result in more adopted a clicks and your hypothesis using a photo of a cat wearing a hat will result in more adopted a clicks you also have your independent variable whether a person clip the adopted a button or not and your independent variable the home page photo you have a control photo no hat or the test photo hat by building up from question two independent variable we know exactly what we're asking and we can already see the shape of our experiment for how to answer our question before we start building our experiment we want to know what our conversion rates look like before changing anything let's take a quick look at that dataset here I'm using the suite of packages called the tiny burst it should be familiar to you from data camps course in the tiny verse from the tiny verse we use the function read CSV from the reader package to load our data here click data if we look at the first few rows we see that we have two columns one visit date which gives the day when someone visited the website and two clicked adopt today which is a one if someone click on the button and is zero if they didn't okay now that we have the basic motivations of a B testing let's practice what we've learned and take another look at our preliminary