1) The chapter's introductory ANALYTICS IN PRACTICE illustration is about Urban Outfitters, Inc. 2) Historical analogy falls under the category of qualitative and judgmental methods 3) Stationary time series data exhibit trend and/or seasonality 4) Cycles are equivalent to seasonality but seasonal cycles tend to repeat more quickly (no longer than every year) 5) Simple moving average forecasting is appropriate for stationary data 6) Simple exponential smoothing is effectively a weighted moving average method 7) The more erratic the time series data, the larger the number of periods one needs for simple moving average forecasting 8) The forecast error in period t is the actual value for period t divided by the forecast for period t 9) The error metric that penalizes large forecast errors the most is the mean absolute deviation (MAD) 10) To apply simple exponential smoothing, you have to specify an initial forecast for the first period 11) The flatter the data, the smaller the smoothing constant one should use for simple exponential smoothing forecasting 12) The simple moving average method yields biased forecasts if the data exhibit a trend 13) If the data exhibit a linear trend, time series regression will outperform simple exponential smoothing 14) In time series regression with seasonality, time is modeled as a dummy variable 15) The Holt-Winter additive model is preferred to the multiplicative mode if from week to week, it is more accurate to measure each day's value as a percentage of the week's daily average instead as an absolute amount above or below the week's daily average 16) To apply simple the Holt-Winters model to data with seasonality, you have to calculate an initial set of seasonal terms before you can begin applying equation 9.18 on textbook page 345 17) Trial and error can be used to find nearly optimum o, B, and y values for the Holt-Winters model 18) Regression with causal variables is based on the same regression principles presented in Chapter 8 19) Causal variables include economic indicators such as consumer confidence 20) For the predictor variables, a given regression model can use EITHER a causal variable OR time but not both