What Are Degrees of Freedom and Effect Sizes? | Crash Course Statistics #28

TL;DR

Degrees of freedom count the independent pieces of information in a dataset, while effect size describes the magnitude of an observed effect relative to random variation. Calculating a mean uses one degree of freedom, so a sample of n observations has n−1 remaining in the example discussed. Together, these concepts give t-tests better context and help distinguish statistical significance from practical importance. Read on to see how sample size and p-values fit into the picture.

Transcript

Hi, I’m Adriene Hill, and welcome back to Crash Course Statistics. It’s great to have a lot of choices. But sometimes we limit our choices in order to do something productive or meaningful. Like being on a team project that needs a writer, director, host, camera person, and boom mic holder. If we have 5 different people who can be on that team, aft... Read More

Key Insights

• Degrees of freedom are the number of independent pieces of information in a dataset.
• Calculating a mean uses up one degree of freedom, reducing the independent information.
• The t-distribution, used in t-tests, changes shape based on degrees of freedom.
• Larger sample sizes provide more accurate estimates, affecting the t-distribution.
• Effect size measures the magnitude of an effect relative to random variation.
• Statistical significance does not always imply practical significance.
• P-values indicate the likelihood of an effect occurring by chance, but need context.
• Effect sizes help determine if observed effects are meaningful in practical terms.

Questions & Answers

Q: What are degrees of freedom in statistics?

Degrees of freedom are the number of independent pieces of information in a dataset. When a calculation such as a mean uses one piece of independent information, the remaining degrees of freedom decrease by one.

Q: Why does calculating a mean use one degree of freedom?

Once the mean and all but one data value are known, the final value is fixed and can be calculated. That last value is therefore no longer free to vary independently, leaving n−1 independent pieces of information.

Q: How do degrees of freedom affect a t-test?

Degrees of freedom determine which t-distribution is used when calculating a p-value for a t-test. They reflect how much independent information remains after calculations such as the mean have been made.

Q: How does sample size affect the t-distribution?

As the sample size n increases, estimates of population parameters such as the mean and standard deviation become more accurate. The resulting t-distribution gets closer to the z-distribution because there is less uncertainty.

Q: Why does the t-distribution have fatter tails than the z-distribution?

A t-statistic uses the sample standard deviation because the population standard deviation is unknown. That extra uncertainty makes extreme t-values slightly more likely, producing a distribution with fatter tails.

Q: What is effect size in statistics?

Effect size measures the magnitude of an observed effect relative to random variation. It adds context by showing whether an effect is large enough to matter in practical terms.

Q: Why should effect size be considered with statistical significance?

A statistically significant result can still have a small effect size and little real-world importance. Considering effect size alongside significance helps distinguish an unlikely-to-be-random finding from a practically meaningful one.

Q: How do p-values and effect sizes differ?

A p-value addresses how likely an observed effect is to occur by chance, while effect size describes the magnitude of that effect. Using both gives a fuller interpretation of whether a result is statistically significant and practically relevant.

Summary & Key Takeaways

• Degrees of freedom are key to understanding statistical models, representing the independent pieces of information available. When calculating a mean, one degree of freedom is used, which affects the t-distribution used in t-tests. Larger sample sizes lead to more accurate estimates and a t-distribution closer to the normal distribution.

• Effect size quantifies the magnitude of an observed effect, providing context to statistical significance. While a result may be statistically significant, the effect size determines its practical relevance. For instance, a small effect size might render a statistically significant result unimportant in real-world applications.

• P-values measure the likelihood of an effect occurring by chance, but they should be evaluated alongside effect sizes to assess practical significance. Degrees of freedom and effect sizes together help in understanding the accuracy and relevance of statistical findings, guiding decisions in various fields.