NAME: BestBuy TYPE: Time series SIZE: 48 observations with 3 variables, and 18 future projections by Best Buy Corporation with 2 variables DESCRIPTIVE ABSTRACT: The dataset bestbuy.day contains monthly data on computer usage (MIPS) and total number of stores from August 1996 to July 2000. Additionally, information on the planned number of stores through December 2001 is available. These data can be used to compare time-series forecasting with trend and seasonality components and causal forecasting based on simple linear regression. The simple linear regression model exhibits unequal error variances, suggesting a transformation of Y. SOURCES: This data is from the Best Buy Co., Inc. Best Buy (NYSE:BBY), headquartered in Eden Prairie, Minnesota, is the largest volume specialty retailer of consumer electronics, personal computers, entertainment software, and appliances. VARIABLE DESCRIPTIONS: Columns 1 - 11 Date dd-mm-yyyy, August 1996 - July 2000 12 - 19 MIPS usage (MIPS are a measure of computing resources) 20 - 30 Number of stores 31 - 44 Date dd-mm-yyyy, July 2000 - December 2001 45 - 53 Planned Number of Stores through December 2001 Values are aligned and delimited by blanks. SPECIAL NOTES: Time-series forecasting is based on MIPS usage and time. Causal forecasting is based on the relationship between MIPS usage and number of stores. STORY BEHIND THE DATA: In August of each year, Best Buy purchases mainframe MIPS (MIPS are a measure of computing resources) in anticipation of the coming holiday season. For planning and budgeting purposes they also wish to forecast the number of MIPS needed the following year. Students are asked to forecast the MIPS needed for December 2000 and December 2001 using the bestbuy.dat data set. Best Buy Corporation actually used this data to predict computer usage in order to budget for and purchase an appropriate amount of computing power. PEDAGOGICAL NOTES: Students can easily understand the seasonality that retail operations experience. Best Buy Corporation has experienced significant growth over the past few years and most students understand that as a firm grows, their need for computing power also increases. Therefore, this data set can be used to demonstrate time-series forecasting with both a trend and seasonality. This data set can also be used to demonstrate causal forecasting based on simple linear regression of computer usage and number of stores. The simple linear regression model exhibits unequal error variances, suggesting a transformation of Y. Finally, a comparison between time-series models and causal models can be discussed with the students. SUBMITTED BY: Julie M. Hays College of Business University of St. Thomas Minneapolis, MN 55403-2005 USA jmhays@stthomas.edu --