 Hello everyone welcome back to learning analytics tools course. In this week we will discuss diagnostic analytics. If you remember this picture from week one we saw that there are four levels of analytics descriptive diagnostic predictive and prescriptive. In the last week we saw descriptive analytics what data to collect how to represent and that is very basics we saw because we do not want to teach a course on designing better visualizations or infographics. But I hope you got the basics of what is descriptive analytics. In this week and coming week we will talk about diagnostic analytics then we will talk about predictive analytics. What is diagnostic analytics? So it is to understand or to identify why this particular thing happened why the students in the year 2018 is not able to score good or why the students in 2016 performed above average why it happened what was the factor imparting it. So to identify that is called diagnostic analytics. And so we saw that diagnostic analytics is identifying why particular thing happened. For example if you have students attendance and midterm marks and their final marks. Let us consider this example for this activity. So we will look at this example often to explain what is diagnostic analytics. For example if you have students attendance midterm marks also the final semester marks the final marks is the dependent variable attendance and midterm of the independent that is x1, x2 final marks is the y1. And we want to know why few students scored less than 40 here the focus is not who scored above 19 instead the focus is on who is scoring less than 40. So let us down based on your experience why would have it happened. So consider you have only these data like midterm marks and attendance. Can you create a hypothesis from this data? And if you have hypothesis how would you like to test it? What are the techniques used to test whether your hypothesis quite or wrong? You consider that you have 1000 students data or 60 students data like number of data is up to your choice think that you have this data and you have x1, x2 and you want to identify the hypothesis the relation between x1, x2 to y1 and come up with hypothesis also come up with a methodology to test it. Please let me the video after you have done your answers. So there are few techniques for diagnostic analytics. One is correlation or clustering, pattern mining or process mining. For example this clustering or process mining can be also converted into a predictor analytics but let us consider this for a diagnostic analytics. So the very basic step is correlation. So what is correlation and what is clustering? We will discuss about correlation in this week. In the coming weeks we will talk about pattern mining process mining then clustering. So how to create hypothesis this is very important because you have data like dependent data also the independent data how do you come up with the hypothesis? The algorithm can be used to test your hypothesis but algorithm will not create hypothesis for you. So algorithm cannot create a research question for you the algorithm can be used to test your research questions or the questions you have. So how to create it? Let us start with the first step like you have to collect data like what data you can collect after you collect data do the descriptive analytics. See visually is there any connection any relationship between this data plot in a bar chart distribution chart and in fact you go for scatter plot see the relationships. Then you have to use your domain knowledge to come up with your hypothesis that is why analytics goes when you apply to a learning analytics it is about you coming up with your domain knowledge to apply on your domain. So your expertise in domain helps you to identify the hypothesis for example we know that the students who are not able to attend a class or were not able to submit assignments on time not able to get a good score in the final exam. If you know that hypothesis and you have a data you want to test it, code and test it. So that kind of hypothesis that kind of such question has to come from the domain expertise. If you are not able to identify the such question or develop an hypothesis from the domain knowledge please read in such papers the recent research papers in the related field and how what kind of questions they are coming up with and how they are creating the questions. You can start with testing the existing questions available on your data then that kind of practice will make you think how to create a new research questions that is how you to read the such papers. So in this video we just got introduction to what is diagnostic analytics and the coming video we will talk about correlation in detail what are types of correlation. Thank you.