 Hi, I'm Dr. Don and I'm going to give you an overview of chi-square test, but let's start with pronunciation of chi-square. It's K-I as in kite, not chi or key square, it's chi-square. You may be familiar with a t-test and we can use a two-sample t-test to see if there's a significant difference in the average of two groups or two levels on a categorical variable on some metric, some measurement. This is an example where we're comparing the average graduation rate of schools in two counties. If you've got more than two groups, more than two levels on your categorical variable, then you can use an ANOVA to compare the averages. Here we're comparing the average salaries of attorneys in a law firm's five offices. We can use a regression to see if two variables are correlated or we can predict one variable from the other. This is a multiple regression and here we've got obesity and poverty as the independent variables predicting a state's cancer rate. What can we use a chi-square for? Let's look at some examples. We've got a manager of a large medical practice and she thinks that patients miss more appointments on Mondays and Fridays and other days. She gathers data from the records and she has the counts of the no-shows that were experienced on five days of the week and if there is nothing unusual then the expected proportion would be 20% per day. She can answer her question using a one-way chi-square goodness of fit. She runs that test and she concludes at the 5% significance level there is no difference in the observed and expected proportions. She also thinks that patients who are unemployed miss more appointments than employed patients. She wants to know is a patient's employment status related to whether a patient makes an appointment is a no-show or cancels in advance. She can use a two-way chi-square test for independence with these data and she runs that test and she determines at the 5% significance level there is a relationship between employment status and appointment outcome. Finally she thinks that there's too many errors being made currently in patient contact information. She gets a sample of patient records for this quarter and last quarter and she wants to answer the question is the error rate in quarter two significantly differ from quarter one and she can answer this question with a two-way chi-square test for homogeneity. So she runs that test and she finds out at the 5% significance level there is no difference in the error rates for these two quarters. So chi-square tests can answer a variety of questions but there are also limitations. You have to have counts for your observed values. You cannot use measurements, something with a decimal, a rate, a percentage. Chi-square is limited to both small samples and very large frequencies. If you've got counts in the thousands then a very small difference is exaggerated. The test will not give you an effect size like a Pearson correlation you get from a relationship. It just says two variables are related or not. And if you're trying to see if order matters, chi-square's not for you. If you were looking at the distributions over four quarters of the year, sliding second quarter into third place wouldn't change the outcome of your chi-square. So what's next? We're going to briefly explain a little more in depth how the chi-square test work and then we will jump into each test and show you how to use technology to solve each one of the three types of chi-square tests. So I hope this helps.