Chapter 1: Introduction

1.1 Forecasting and Data

What is a Forecast?
Cross-Sectional Data
Time Series Data
             Trend,   Cycle,    Seasonal Variation,    Irregular Variation


1.2 Forecasting Methods
Qualitative Forecasting Methods  ("Judgmental forecasting")
   
Subjective Curve Fitting
                           
Product Life Cycle
Delphi Method
Time Independent Technological  Comparison

Quantitative Forecasting Methods

Univariate Forecasting Models: Everything is a function ot time and/or past y
Causal Forecasting Models: The observedvalue of x "causes" us to believe that y will be f(x)
                     For example, seeing smoke now (x) causes us to think we will soon see fire (y).
             "Leading Indicator Models" is a less misleading name for the same thing.


1.3 Errors in Forecasting                    (Statistics is the Science of Error)
Types of Forecasts

Point Forecasts
Prediction Interval Forecasts   
                ("Probabilistic Forecasts diagram"
("Future forecast" really is presnet forecast of future value!)


Measuring Forecast Errors

Forecast Error = actual - forecasted.
                           a Positive error means the forecast was Too Low!   (silly, isn't it?)
Bias:  It's OK to be too high sometimes and too low other times as long as it averages out.

Absolute Deviation
Mean Absolute Deviation: Being off by 3, plus or minus, is 3 tomes as bad as being off by 1, pluis or minus

Squared Error
Mean Squared Error
: Being off by 3, plus or minus, is 9 tomes as bad as being off by 1, pluis or minus
   
Absolute Percentage Error
Mean Absolute Percentage Error


1.4 Choosing a Forecasting Technique
1. The Time Frame
                       Forecasting one period ahead
                       Forecasting several periods ahead
                       Guessing about the distant future
2. The Pattern of the Data
                       Based on business knowledge, not statistical fishing!
3. The Cost of Forecasting
                      Developing the Model
                      Operating the Model
                                   For quantitative methods, nearly all the cost is obtaining valid data.
                                   For qualitative methods, it's paying experts' salaries or consulting fees.
4. The Accuracy Desired
                       Based on the cost of error and the accuracy of available approaches
5. The Availability of the Data
                       What's possible to obtain? How much will it cost (see above)?
                        How reliable is it?  What will it cost to make it more reliable?
6. The Ease of Operation and Understanding
                        Data Collection and Data Entry once again
                        Why should I bet my career on your forecast?