Questions (AllRoundBest Tutor)

Week Four Listen To Me First

Speakers: Narrator, SME

Narrator: Welcome to the Week Four podcast for CJA/335. This week, we’ll discuss statistical significance and tests.

What does it mean to have statistical significance?

SME: First, statistical significance should not be confused with practical significance. Significance is traditionally interpreted as important meaning, whereas in statistics, significance really means that something is not by chance and is probably true. Statistical significance is measured mathematically. With statistical significance, a mathematically small value can be highly significant if the sample size is large enough. Finding statistical significance means an effect is "real" and not due to chance or coincidence. Mathematically, statistical significance can come down to an integer with two or more decimal places, such as 0.01, 0.001, and so on. Statistical significance is both about the probability and the strength of the relationship.

Narrator: When do we need to conduct significance tests on data?

SME: You usually conduct significance tests in instances when you want to determine if the evidence or data can refute the null hypothesis. The null hypothesis is that the correlation being hypothesized is not present. If the probability is less than 0.01, you have determined that the null hypothesis is likely false. Another way to say it is that you have a less than one percent probability that a dependency, effect, or correlation is by chance.

Narrator: And what key point should students focus on this week?

SME: Statistical significance and testing is in large part built on the knowledge and skills of correlations and data dependency. It has a lot to do with hypothesis testing and is integral to verifying and validating a research design and interpreting the data results. Students should explore the different significance tests and why one is used over another in different cases. Lastly, please don't forget, significance is an estimation and is not absolute.