QMETH: 520 (W) & 530 (SP) MBA Orientation April 1, 2005
Forecasting-Two Approaches Forecast = f (Data Model Software) f (Data, Judgment) Survey Data (QMETH520) Past Data (QMETH530)
Survey Data (520) – An Example
“Variety” of The Dependent Variable Problem that 520 focuses on Dependent Variable, Y Quantitative (Amount) Qualitative (Choice) Continuous (measurement) Discrete (counting) Ordinal Nominal
Management Applications Health care: Reasonable cost estimation Finance: Credit scoring Marketing: Customer targeting Marketing: Effect of promotion for purchase Human resource: Performance prediction, Determination of fair salary
Past Data (530) Time Series –Variables observed in regular frequencies –Daily, Weekly, Monthly, Quarterly, Yearly, etc.
“Variety” of Dynamics Problem that 530 focuses on Trend Seasonality Cycle Unequal variance Example 1
Cycle Example 2:
Conditional Unequal Variance Example 3
Management Applications Sales forecasting Forecasting of macro economics variables Forecasting of the risk of financial returns Forecasting the effect of the system interventions
Model Based Forecasting Forecast = f (Data, Model, Software) Tremendous technological progress in: –Developing standard forecast models –Developing efficient, easy to use software
Modeling Process We do not reinvent a new wheel We “match data” with a “standard model” Data Standard Forecasting Models
Role of Software SPSS for 520 and Eviews for 530 Graphical Display of Data –Guiding the choice of models Model Fitting & Evaluation –Fitting standard models supported in the software –Evaluating adequacy of the models after fitting Forecast –Computing forecasts
Implications of Using Standard Models Low Learning Cost –Almost all in a family of “regression” –Democratization of forecasting technology Transparency of forecasting process Identify the weaknesses of the forecast process