Exploring the power calculation

As discussed, power is the probability of rejecting the null hypothesis when the alternative hypothesis is true. Here you will explore some properties of the power function and see how it relates to sample size among other parameters. The get_power() function has been included and takes the following arguments in the listed order n for sample size, p1 as the baseline value, p2 as the value with lift included, and cl as the confidence level.

Cet exercice fait partie du cours

Customer Analytics and A/B Testing in Python

Afficher le cours

Instructions

Calculate the power using n = 1000 and n = 2000 in that order, along with the pre-loaded parameters, p1, p2, and cl.
Using the variable n1 for the sample size, find the power with a confidence level of cl = 0.8 and cl = 0.95 in that order.
Hit 'Submit Answer' to compare the ratios. Which change has the bigger impact, increasing the confidence level or the sample size?

Exercice interactif pratique

Essayez cet exercice en complétant cet exemple de code.

# Look at the impact of sample size increase on power
n_param_one = get_power(n=____, p1=p1, p2=p2, cl=cl)
n_param_two = get_power(n=____, p1=____, p2=____, cl=____)

# Look at the impact of confidence level increase on power
alpha_param_one = get_power(n=n1, p1=p1, p2=p2, cl=____)
alpha_param_two = get_power(n=____, p1=____, p2=____, cl=____)
    
# Compare the ratios
print(n_param_two / n_param_one)
print(alpha_param_one / alpha_param_two)

Modifier et exécuter le code

Cet exercice fait partie du cours

Customer Analytics and A/B Testing in Python

IntermédiaireNiveau de compétence

4.8+

Commencer le cours gratuitement

This chapter provides a brief introduction to the content that will be covered throughout the course before transitioning into a discussion of Key Performance Indicators or KPIs. You'll learn how to identify and define meaningful KPIs through a combination of critical thinking and leveraging Python tools. These techniques are all presented in a highly practical and generalizable way. Ultimately these topics serve as the core foundation for the A/B testing discussion that follows.

Exercise 1: Course introduction and overview Exercise 2: Understanding the key components of an A/B test Exercise 3: Defining meaningful KPIs Exercise 4: Identifying and understanding KPIs Exercise 5: Loading & examining our data Exercise 6: Merging on different sets of fields Exercise 7: Exploratory analysis of KPIs Exercise 8: Practicing aggregations Exercise 9: Grouping & aggregating Exercise 10: Calculating KPIs - a practical example Exercise 11: Calculating KPIs Exercise 12: Average purchase price by cohort

This chapter teaches you how to visualize, manipulate, and explore KPIs as they change over time. Through a variety of examples, you'll learn how to work with datetime objects to calculate metrics per unit time. Then we move to the techniques for how to graph different segments of data, and apply various smoothing functions to reveal hidden trends. Finally we walk through a complete example of how to pinpoint issues through exploratory data analysis of customer data. Throughout this chapter various functions are introduced and explained in a highly generalizable way.

Exercise 1: Working with time series data in pandas Exercise 2: Parsing dates Exercise 3: Creating time series graphs with Matplotlib Exercise 4: Plotting time series data Exercise 5: Pivoting our data Exercise 6: Examining the different cohorts Exercise 7: Understanding and visualizing trends Exercise 8: Seasonality and moving averages Exercise 9: Exponential rolling average & over/under smoothing Exercise 10: Events and releases Exercise 11: Visualizing user spending Exercise 12: Looking more closely at revenue

In this chapter you will dive fully into A/B testing. You will learn the mathematics and knowledge needed to design and successfully plan an A/B test from determining an experimental unit to finding how large a sample size is needed. Accompanying this will be an introduction to the functions and code needed to calculate the various quantities associated with a statistical test of this type.

Exercise 1: Introduction to A/B testing Exercise 2: Good applications of A/B testing Exercise 3: General properties of an A/B Test Exercise 4: A/B test generalizability Exercise 5: Initial A/B test design Exercise 6: Experimental units: Revenue per user day Exercise 7: Preparing to run an A/B test Exercise 8: Conversion rate sensitivities Exercise 9: Sensitivity Exercise 10: Standard error Exercise 11: Calculating sample size Exercise 12: Exploring the power calculation

Exercice en cours

Exercise 13: Calculating the sample size

After running an A/B test, you must analyze the data and then effectively communicate the results. This chapter begins by interleaving the theory of statistical significance and confidence intervals with the tools you need to calculate them yourself from the data. Next we discuss how to effectively visualize and communicate these results. This chapter is the culmination of all the knowledge built over the entire course.

Exercise 1: Analyzing the A/B test results Exercise 2: Confirming our test results Exercise 3: Thinking critically about p-values Exercise 4: Understanding statistical significance Exercise 5: Intuition behind statistical significance Exercise 6: Checking for statistical significance Exercise 7: Understanding confidence intervals Exercise 8: Calculating confidence intervals Exercise 9: Interpreting your test results Exercise 10: Plotting the distribution Exercise 11: Plotting the difference distribution Exercise 12: Finale