Forecasting Models in Excel

Forecasting Models in Excel

Time Series Models

  • Time series models are often used in forecasting.

  • Excel can generate a time series model automatically.

Model Components

  • Alpha (α):

    • Represents the level or intercept of the model.

  • Beta (β):

    • Pertains to the trend component of the model.

    • This is analogous to a coefficient multiplied by time, denoted as the term multiplied by time, (t).

  • Gamma (γ):

    • Represents the seasonality within the model.

    • An example is the increased sales seen in the fall season.

Optimization Process

  • The model is generated, and it seeks to achieve the best fit by optimizing parameters α, β, and γ.

  • The aim is to minimize the Sum of Squared Errors (SSE) or Group Mean Squared Error (GMSE).

  • By minimizing these errors, it determines the most accurate model representation.

  • Optional: For students interested, comparing Excel's output with results from lab work is encouraged to see differences in forecasting accuracy.

Types of Forecasting Models

1. Regression/Econometric Models

  • These models use explanatory (independent) variables to predict a dependent variable.

  • Example:

    • If the price changes, how does that affect quantity sold?

  • A structural relationship must be established between variables to justify predictions.

  • Previous lab work has involved time series modeling and regression analysis to improve forecast accuracy.

2. Barometric Models

  • These models identify patterns or indicators that suggest future events.

  • Example:

    • A drop in factory orders could signal upcoming unemployment increases.

  • Look for relationships in data that can predict future occurrences.

  • Hypothetical scenario connecting animal behavior successfully predicting earthquakes illustrates the application of barometric models.

  • Barometer analogy:

    • A barometer measures air pressure; a significant drop may predict rainfall.

    • This correlates with the idea of volumentric forecasting based on observed changes in indicators.

Case Study: Groundhog Day

Groundhog as Weather Indicator

  • Punxsutawney Phil's emergence is used as a cultural metaphor for predicting winter length and symbolizes forecasting practices in a humorous sense.

  • Reference to the movie "Groundhog Day" (1993) provides a pop culture connection to the theme of prediction.

Practical Data Analysis with Google Trends

  • Data from Google Trends can also be used to investigate potential correlations and forecast future events.

Example Data:

  • The presenter gathered Google Trends data on sports such as:

    • Key Jumping

    • Curling

    • Hockey

  • The trends reveal varying search volumes correlated with events like the Olympics (indicated by the blue line representing curling).

Discussion on Nordic Skiing

  • Mention of athletes such as Jesse Diggins who became prominent during recent Olympic events (10 km freestyle ski event).

  • Discussion of athlete-related data connections to public interest and any potential relationship with organizations like the EMILY program.

Summary of Lab Work

  • Review of methods used in previous labs:

    • Time trend analysis

    • Polynomial regression (e.g., (t^2)), which typically shows a flat line.

    • Lagged price variables: (price{t-1}), (price{t-2}).

    • Moving average methods for trend analysis.

  • Contrast to intuitive predictions (referred to as "animal spirits").

  • Discussed the attempt to forecast all of 2025 using data available only until December 2024 without any significant seasonal indicators leading to a linear prediction.

  • Encouragement for students to think critically about the differences in data analysis methods and their forecasting accuracy.

Questions and Closing Remarks

  • Open floor for questions; prizes will be given for participation and engagement in discussions.

  • Encouragement for students to explore the material further and consider how various forecasting methods can be applied in practical settings.