Mr. Palsson plans to buy a MacBook Pro which can hold the 6GB of RAM (he has a 4GB chip laying around, so he doesn't care what RAM is actually in the computer, as long as the computer can handle 6GB), swap the computers' hard drives, and then sell his current computer, with the hard drive from the upgrade computer. To clarify: First he will buy computer off eBay that meets his upgrading needs. He will then swap the hard drives. Finally he will sell his old computer with the hard drive from the computer he just bought. This will save him a lot of installing software hours!
Mr. Palsson is married to an operations research specialist, who quickly realizes that regression can help with his decision problem. At the same time the OR wiz realizes (although she decides not to tell him) that Mr. Palsson may not have collected the best data - but it would have to do. Some of it may even have to be thrown out - but such are the realities of Data and Modeling.
Using the data on successful transactions (actions that ended in a sale), build a regression model(s) to assist Mr. Palsson with upgrading his computer.
Word of advice! Be selective about how you model your variables. Mr. Palsson has not provided us with too many data points, so the data cannot support fitting over 30 variables - so I suggest, for example, not creating a dummy variable for each day of the week!1 Use your judgment and be selective. In addition, with few data, but many variables, pay attention, not only to R2 but adjusted R2 as well. In the end you should have a model were the signs of the coefficients make intuitive sense (for the most part)2.
Using data on successful (result=1) auction style (type equals 1, 2, 5 or 6) transactions, build a model that predicts the sale prices of a MacBook Pro based on the data Mr. Palsson has collected. Provide a printout of your model and label it as Exhibit 3-a. (Note: there are multiple possible successful models here. Please only include information about ONE model, the one you deem best to use).
Explain or provide justification for data modeling and/or nonlinear transformations you use in your model in the space below. If you want to submit supporting Exhibits, please label them etc. and clearly list them in the space provided below. The Exhibits should be self-explanatory - the grader should not need to guess what they mean.
1 Perhaps a good rule of thumb is for your dummy variables to have no less than five observations of both values (zero or one)
2 This question is not meant to be a tricky exam problem addressing nonlinear transformations.
Please answer the following questions based on your model in part a) in the space provided.
Based on your model what can mr. Palsson expect to pay for a upgrade computer if he buys it in an auction style listing (assume he buys one with a 250GB hard drive, in a good condition)?
Based on your model what can he expect to get for his computer if he lists it in a auction style format (assume he sells his computer with a 250GB hard drive)?
What is the right selling and buying strategy for Mr. Palsson if he utilizes auction style listings?
Using data on transactions that finished with buy-it now (result=2), use your model from a) to analyze how the ending prices in a buy it now format compare to auction style (HINT: how much would you expect these computers to sell for if listed auction style). Create an exhibit that summarizes your findings and label it Exhibit 3-b.
Should Mr. Palsson utilize the "buy it now" format for either buying or selling a computer? Briefly explain your reasoning.
Assume the upgrade computer mr. Palsson buys has a 250GB hard drive. Using your analysis so far (and any additional analysis you chose to conduct), is it worth it for Mr. Palsson to buy a new 500GB hard drive for $90 to insert into his old computer before selling it, rather than to just put in the 250GB hard drive from the upgrade computer? Briefly explain