We're now going to start discussing more complex types of analysis beginning with association analysis in comparing groups of people. In this lesson, we'll work through an example to determine if a company should consider changing the advertising strategy for its app. So, let's start with association analysis, which helps us determine whether or not differences between groups exist. So far we've used states that involve one homogeneous groups. In marketing however, being able to measure access, whether there are differences between groups, is important for segmentation and positioning purposes, two very important strategic process in marketing. A common type of association analysis is to determine whether certain groups or types of customers behave differently or not. Consider the following question: Do men drink more soft drinks than women? If I were to work for a company selling soft drinks, it will be interesting to know if one group has a higher consumption than another. That could allow me, for example, to define different marketing campaigns. To answer questions such as this a test exist. Consider the following example, imagine you're working for a food delivery mobile application, and that this company just finished an advertising campaign. The advertising campaign was on both offline, let's say, on TV or radio, and online, let's say, through display advertisements. Whenever a person downloaded the app, he or she was asked to answer a short survey, to answer whether or not she remembers seeing some of the company's advertisements in the week prior to downloading the app. Some of these people then decided to subscribe to the app, for which they paid a monthly fee to have food delivered to the door, and some people decided not to. You can probably see that for the company, understanding what drives people from simply downloading the app to pay for a service, is something that is very important. So, what we want to understand is whether or not there are differences in how people join, or do not join based on that exposures. When we look at these data we see that when people subscribed, 45 people remember seeing the ads for the app offline, and 89 of these people remember seeing the ad online. For those that chose not to subscribe,15 people remember seeing the ad offline and 20 people online. So, the question here is, are there any significant differences between the two groups? We are going to assume again 95% confidence level. If there are differences between the two groups, this could be useful for managing ad budgets. Recall that the null hypothesis is going to be rejected, if the Z-value that we are going to compute is below [inaudible] threshold value. Again, the null hypothesis in this case would be, that the two groups are going to be equal, that the proportion will be equal across the two groups. And the alternative hypothesis is that those differences exist, so the proportions should be different. At the 95% confidence level, the critical value of Z is equal to 1.96 and the formula for the simple statistic is going to be equal to proportion one minus proportion two divided by the standard deviation of the difference of the two proportions which itself is equal to the square root of (p1×q1÷n1) + (p2×q2÷n2). Plugging the numbers in the formula, we see the Z-value for online ads is 0.48 which is below 1.96. So, in the case of offline ads, we can say that there is not enough support to reject the null hypothesis with respect to differences between the people who join and people who didn't join. Therefore, we can assume that offline ads do not affect these groups differently. On the other hand, when you compute the statistics for online ads, we see that the Z-value is above the critical value 1.96 number. And in this case, we can conclude that we should reject the null hypothesis, which means that there are differences between the two groups in how they subscribe to the service.