Let's talk about the Kruskal-Wallis Test. After this video, you will be able to perform this test and you will be able to interpret the results. But first, let's look at when to use the Kruskal-Wallis test. Our tree diagram shows that it is an alternative to the ANOVA technique. And hence, the Kruskal-Wallis test is used to analyze the relationship between a numerical CTQ and a categorical influence factor. The Kruskal-Wallis test is a non-parametric test. This means that there is no specific distribution assumed on the residuals. Remember that the ANOVA was based on the assumption that the residuals are normally distributed, and that the residuals contain no irregularities such as outliers. Of course, if you let go of assumptions, the Kruskal-Wallis test will be less powerful than the ANOVA analysis. This means that you basically need more data to show the same difference. Okay, now let's take a look at an example. Consider a call center. Some of the employees received a training while some of the others did not. You want to know whether this training has any effect on the total handling time, that is, the total time it takes an employee to handle an incoming call. You have measured the total handling time for some calls, and have recorded if the employee handling the call has received this training. Remember that you should always start by first identifying your Y variable. That is the total handling time in this example, and it is numerical. Next, identify your influence factor X. That is training or not, and it is categorical. Now, the tree diagram will tell you to analyze this relationship between total handling time in training using an ANOVA analysis. Performing this analysis, you will get in your third step, this four in one plot. As you can see, the residuals are not normally distributed. Furthermore, the time graph also shows some irregularities, hence we need to apply an alternative analysis technique. In this case, we will apply the Kruskal-Wallis test. The data that is gathered for the call center looks like this. Now, please pause the video, load this data into Minitab before continuing. After loading your data into Minitab, it should look like this, with Total Handling Time in the first column and Training in the second column. Now I will show you how to perform a Kruskal-Wallis analysis using Minitab. Of course, as it is a statistical test, you will find it under the menu Stat. Remember that the Kruskal-Wallis test is a nonparametric test so we go to the menu Nonparametrics. Here you will find the Kruskal-Wallis analysis. Now, Minitab asks us for the response, which is your y or your CDQ. And that's the total handling time. Next, we have to fill in the factor, which is the training. And okay. Now, let's study the Minitab output which you can find in the Session window. The Kruskal-Wallis analysis is based on medians which are reported. The median handling time for employees with training is 191 seconds. And the median handling time for employees without training is 246 seconds. So it took people without training a lot more time to handle a phone call than it did for people that received the training. The difference is nearly a minute. The p-value is also given. It is lower than 0.05. So the difference in medians that we found is statistically significant. It can be concluded that the training is an effective method to lower the total handling time by 55 seconds. Now, let's summarize. You learned to perform a Kruskal-Wallis test. This test should be performed when the underlying assumptions of the ANOVA analysis are not met. To interpret the output, you have to look at the medians for each group and of course, add the P-value to see if this difference is statistically significant.