In this module we'll be looking at explanatory modeling. Basically, in the previous module we saw that the model is just one part of this entire business analytics. Here we're going to look at a particular modeling method, which all of us have learned in some time or other in our courses called regression, but we will use it in the context of. We're going to look at four segments in this module. The first is, what is an explanatory model. And so let's go through a few which you're familiar with and ask ourselves, what's common about them. Second, this depends on the kind of model you're going to use. But as we're going to use regression, we're going to start looking at how do you even go about developing a model and developing our confidence and using that model? Once you're confident, and confidence comes from understanding the output of regression, we can then start saying, are we happy with this model? Can we make it better? Is that all the things that we can do to make it better? Notice that the steps 2, 3 and 4 are common to most of the tools that we'll be using in this course. So let's take a few examples. So if you bought a house, now buying houses is a big decision. You may have to pay back for it over 30 years or 15 years. So people are very careful and do a lot of analysis when they buy a house. They try to see how many bedrooms it has, how big is the lot. How are the schools? What zone, what's the price of taxes you pay? What are the prices of homes sold nearby, right? And so what we want to understand is what explains the price of house, sometimes there are real surprises. Hey, why is this costing so much? Or sometimes they say hey, this looks too low. There must be something wrong with it. But in these situations, we're trying to build in our mind an explanatory model. Many of you have your cell phones. I know, sometimes that work, sometimes they don't. But sometimes they work well, sometimes they don't work. And we know that in the desert it might not work on a mountaintop it may not work, but it may not work near a high tension tower. And so what determines the performance of a cellular phone? You would like to explain that. Now think that you want to open a store within a mall and you have lots of options. And you want to know, maybe it's a store which is selling fashion goods and you want to know how much will I sell? What goes through your mind is who comes to this mall? What is the average sales per store in this mall, right? And how many competitors do we have? And these are some of the things which might explain the sales in a store. Going on, some may lead to a new product, but I'll skip that. But the next one is simple, you go to a bank for a loan, and sometimes you're really surprised at the number of questions they ask you. If they like you, why can't they hand me the money? But they want to figure it out will you repay that loan and will you repay in time? And will you pay sooner than it is due or late? And they really want to know whether you're a good customer and what explains being a good customer? If you are manufacturer and you got a bunch of suppliers, and I always wondered why is supplier A more dependable than B? Does it depend on the past? Is the supplier having any cash flow problems? How have his deliveries done in the past? Has the rejection level been high? This are some of the things which might explain how good a supplier performs, and whether the performance depends maybe on weather, on geographical conditions,on political issues. Finally, you got a bunch of stores, maybe you hear about store closings in Champaign, I seem to hear about them more often than not. Why does a store close on the main street of Urbana, but maybe survives in a mall? So what determines success of a store or a branch of a bank in a small town like Champaign? In all these questions, what we're really asking ourselves is, what does our experience tell us, right? We're trying to say based on how we have interacted and we've seen in the past, can we explain the success or failure or price instead of our experience, let's substitute data, right? So instead of saying, I know my experience says these are the things that matter, but now why don't we collect the things that matter and then put it in a model and use it to explain the price, explain the performance, explain the stores. So what this step really, it's very important to understand is we are translating our prior knowledge and our interviews with a lot of people into deciding what kind of data might explain the phenomenon we are interested in.
