Overview of Forecasting Two Approaches to Forecasting Using
- Slides: 20
Overview of Forecasting
Two Approaches to Forecasting Using Survey Data (QMETH 520) Model Based Using Past Data Forecasting Methods (QMETH 530) Judgmental (NB: Ch. 11)
Past Data • Time Series – Variables observed in equal time space – Frequency • Daily, Weekly, Monthly, Quarterly, Yearly, etc.
Steps for Statistical Forecasting 1. Determine the variable(s) 2. Collect data • Frequency • Range 3. Develop a forecasting model (DGP) 4. Determine the forecast horizon 5. Determine the forecast statement
Data Sources • Public – Links to several data sources available on the Courses Web • Private
Forecast • Horizon – h step ahead – Short run – Long run h small h large • Statement – Point (unbiased and small se) – Interval (confidence level) – Density
Loss Functions • L(e=y – pred_y) L 0 L e 0 e
Example Variable: Japanese Yen per US Dollar Frequency: Monthly Data Range: 1980: 1 – 2000: 3 Forecast Horizon: 2000: 4 - 2002: 7
Forecasting Model • Statistical (scientific) forecast uses a “model” for determining the forecast statement. • Model = Data Generating Process (DGP)
Standard Forecasting Models • See the list in the syllabus
Modeling Process • We do not reinvent a new wheel • We “match data” with a “standard model” Data Standard Forecasting Models
Importance of Coverage • Merit in learning a variety of forecasting models – Rather than mastering a one particular model • For time series data – To cope with different types of “dynamics” • Survey data – To cope with different types of “variables”
Variety of Dynamics • Data = Trend + Season + Cycle + Irregular • Irregular – Equal Variance – Unequal Variance
Implications of Using Standard Models • Democratization of forecasting technology • Transparency of forecasting process • Identify the weaknesses of modeling – Imperfect model – Not enough observations – Contaminated data
Role of Software • Graphical display of data – Guiding the choice of models • Data Analysis: Matching Process – Fitting standard models supported in the software – Testing the adequacy of the models after fitting • Forecast – Computing forecasts
Forecasting in Action • Operations Planning and Control – Inventory management – sales force management – production planning, etc. • Marketing – pricing decisions – advertisement expenditure decisions
Forecasting in Action - cont. • Economics – macroeconomics variables – business cycles • Business and Government Budgeting – revenue forecasting – expenditure forecasting • Demography – population – immigration, emigration – incidence rate
Forecasting in Action - cont. • Human Resource Management – employee performance • Risk Management – credit scoring • Financial Speculation – stock returns – interest rates – exchange rates
Models Components Forecasting Model Trend Fixed vs. Variable Season Fixed vs. Variables Cycle ARMA Irregular Random / GARCH
Statistical Thinking for Management World Data Information about a few customers, incidents Statistics not used Represent many others Identify the relevant Process Statistical methods needed
- What are the two types of forecasting
- The main approaches to forecasting exchange rates are
- Corporate governance is which approach
- The two dominant agile approaches are
- Forecast horizon
- Using system using system.collections.generic
- Accumulator ac
- A simile compares two things
- Use the following pictures to compare
- Simile definition
- A comparison of two unlike things using like or as
- To find the hcf of two numbers using activity method
- Using graphs to relate two quantities
- Relative motion of two particles using translating axes
- Polynomial multiplication using linked list in c
- Literary techniques comparison
- How to measure 45 minutes using two identical wires
- Join the sentences using relative pronouns
- Comparing two things without using like or as
- Using graphs to relate two quantities
- Factor of safety formula yield strength