Testing for normality

A powerful suite of statistical tools, which includes several common hypothesis tests, depends on the assumption that the underlying data is normally distributed. While a histogram can hint at whether the data is approximately normally distributed, various hypothesis tests allow us to test this assumption directly. Moreover, histograms can be very sensitive to the number of bins, especially when the sample sizes are small.

In this exercise you'll work with salary data from employees for the City of Austin in salary_df. In particular you will be working with Hispanic firefighters. You'll analyze if their years of employment are approximately normally distributed using the Anderson-Darling hypothesis test.

This exercise is part of the course

Foundations of Inference in Python

Exercise instructions

Plot a histogram showing the Years of Employment for the employees.
Conduct an Anderson-Darling test for normality to see if Years of Employment is approximately normally distributed.
Find which critical_values the test statistic is greater than.
Print the significance_level(s) at which the null hypothesis would be rejected.

Hands-on interactive exercise

Have a go at this exercise by completing this sample code.

# Plot a histogram of the employees' "Years of Employment"
____.plot(kind="____")
plt.show()

# Conduct an Anderson-Darling test using the years of employment from salary_df
result = stats.____(____)

# Print which critical values the test statistic is greater than the critical values
print(result.____ > result.____)

# Print the significance levels at which the null hypothesis is rejected
print(result.____[result.____ > result.____])

Edit and Run Code

This exercise is part of the course

Foundations of Inference in Python

AdvancedSkill Level

4.8+

Start Course for Free

In this chapter, we'll explore the relationship between samples and statistically justifiable conclusions. Choosing a sample is the basis of making sound statistical decisions, and we’ll explore how the choice of a sample affects the outcome of your inference.

Exercise 1: Statistical inference and random sampling Exercise 2: Sampling and point estimates Exercise 3: Repeated sampling, point estimates and inference Exercise 4: Sampling and bias Exercise 5: Visualizing samples Exercise 6: Inference and bias Exercise 7: Confidence intervals and sampling Exercise 8: Normal sampling distributions Exercise 9: Calculating confidence intervals Exercise 10: Drawing conclusions from samples

Learn all about applying normality tests, correlation tests, and parametric and non-parametric tests for sound inference. Hypothesis tests are tools, and choosing the right tool for the job is critical for statistical decision-making. While you may be familiar with some of these tests in introductory courses, you'll go deeper to enhance your inferential toolkit in this chapter.

Exercise 1: Normality tests Exercise 2: Testing for normality

Current Exercise

Exercise 3: Distribution of errors Exercise 4: Fitting a normal distribution Exercise 5: Correlation tests Exercise 6: Testing for correlation Exercise 7: Autocorrelation Exercise 8: Explained variance Exercise 9: Parametric tests Exercise 10: Equal variance Exercise 11: Normality of groups Exercise 12: ANOVA Exercise 13: Non-parametric tests Exercise 14: Comparing rankings Exercise 15: Comparing medians

In this chapter, you'll measure and interpret effect size in various situations, encounter the multiple comparisons problem, and explore the power of a test in depth. While p-values tell you if a significant effect is present, they don't tell you how strong that effect is. Effect size measures how strong an effect a treatment has. Master the factors underpinning effect size in this chapter.

Exercise 1: Effect size Exercise 2: Effect size for means Exercise 3: Effect size for correlations Exercise 4: Effect size for categorical variables Exercise 5: Multiple comparisons and corrections Exercise 6: Multiple comparisons problem Exercise 7: Bonferonni-Holm correction Exercise 8: Power of a test Exercise 9: What is power anyway?Exercise 10: Power for experimental design Exercise 11: Computing power and sample sizes

You’ll expand your inferential statistics toolkit further with a look at bootstrapping, permutation tests, and methods of combining evidence from p-values. Bootstrapping will provide you with a first look at statistical simulation. In the lesson meta-analysis, you’ll learn all about combining results from multiple studies. You’ll end with a look at permutation tests, a powerful and flexible non-parametric statistical tool.

Exercise 1: Bootstrapping Exercise 2: Bootstrap confidence intervals Exercise 3: Bootstrapping vs. normality Exercise 4: Combining evidence from p-values Exercise 5: Fisher's method in SciPy Exercise 6: Inference using Fisher's method Exercise 7: Summarizing Fisher's method Exercise 8: Permutation tests Exercise 9: Permutation tests for correlations Exercise 10: Permutation tests and bootstrapping Exercise 11: Analyzing skewed data with a permutation test Exercise 12: Course wrap-up video