7.2 Transfer Function Models

Transfer function models are like secret agents, uncovering hidden connections between variables over time. They reveal how one thing affects another, considering delays and ripple effects. It's like solving a puzzle of cause and effect.

These models are super useful in economics, marketing, and environmental studies. They help us understand complex relationships, like how oil prices impact airline stocks or how advertising influences sales. It's all about connecting the dots in a time-based world.

Transfer function models

Concept and structure

Top images from around the web for Concept and structure

• 

  Time Series Analysis View original

  Is this image relevant?

• 

  Frontiers | Short-Term Wind Power Forecasting Using Mixed Input Feature-Based Cascade-connected ... View original

  Is this image relevant?

• 

  Types of Regression View original

  Is this image relevant?

• 

  Time Series Analysis View original

  Is this image relevant?

• 

  Frontiers | Short-Term Wind Power Forecasting Using Mixed Input Feature-Based Cascade-connected ... View original

  Is this image relevant?

1 of 3

Top images from around the web for Concept and structure

• 

  Time Series Analysis View original

  Is this image relevant?

• 

  Frontiers | Short-Term Wind Power Forecasting Using Mixed Input Feature-Based Cascade-connected ... View original

  Is this image relevant?

• 

  Types of Regression View original

  Is this image relevant?

• 

  Time Series Analysis View original

  Is this image relevant?

• 

  Frontiers | Short-Term Wind Power Forecasting Using Mixed Input Feature-Based Cascade-connected ... View original

  Is this image relevant?

1 of 3

• Transfer function models capture the dynamic relationship between a dependent variable (output series) and one or more independent variables (input series) over time
• The model structure consists of:
  • Output series (dependent variable)
  • One or more input series (independent variables) related through a transfer function
• The transfer function describes how the input series affects the output series, considering:
  • The dynamic nature of the relationship
  • Potential lags between input and output
• Transfer function models incorporate both current and past values of the input series to explain the behavior of the output series
• The model can include multiple input series, each with its own transfer function, to capture the combined effect on the output series
• Transfer function models also include an error term, which represents the unexplained variation in the output series not accounted for by the input series (residuals)

Applications and examples

• Transfer function models are commonly used in various fields, such as:
  • Economics (modeling the impact of interest rates on GDP growth)
  • Marketing (analyzing the effect of advertising expenditure on sales)
  • Environmental studies (examining the relationship between pollutant levels and health outcomes)
• Example: A transfer function model can be used to study the impact of oil prices (input series) on airline stock prices (output series), considering the lagged effects and potential nonlinearities in the relationship

Transfer function models with exogenous variables

Developing transfer function models

• Exogenous variables are independent variables determined outside the system being modeled and used as inputs in transfer function models
• The first step in developing a transfer function model is to identify the relevant exogenous variables that significantly impact the output series
• The relationship between exogenous variables and the output series should be examined using cross-correlation functions (CCFs) to determine the appropriate lag structure
• The transfer function for each exogenous variable is specified based on:
  • The observed cross-correlation pattern
  • The expected dynamic relationship
• The transfer function can be represented using a rational polynomial, which consists of:
  • A numerator polynomial (captures immediate and lagged effects)
  • A denominator polynomial (accounts for the persistence of the effect over time)

Estimating and specifying transfer functions

• The model estimation process involves determining the coefficients of the transfer functions and the error term using techniques such as:
  • Least squares estimation
  • Maximum likelihood estimation
• The numerator polynomial of the transfer function captures:
  • The immediate impact of the exogenous variable on the output series
  • The lagged effects of the exogenous variable on the output series
• The denominator polynomial of the transfer function accounts for the persistence or decay of the effect over time
• The order of the numerator and denominator polynomials is determined based on the observed cross-correlation pattern and the model's goodness-of-fit
• Example: In a transfer function model for sales (output series) and advertising expenditure (input series), the numerator polynomial may capture the immediate and delayed impact of advertising on sales, while the denominator polynomial may represent the diminishing returns or saturation effect of advertising over time

Parameters and components of transfer function models

Interpreting transfer function coefficients

• The coefficients of the transfer function numerator polynomial represent the magnitude and direction of the impact of the exogenous variable on the output series
• Positive coefficients indicate a positive relationship (an increase in the exogenous variable leads to an increase in the output series)
• Negative coefficients indicate an inverse relationship (an increase in the exogenous variable leads to a decrease in the output series)
• The lag structure of the numerator polynomial determines the timing of the impact, with higher-order lags indicating delayed effects

Understanding the error term and its properties

• The coefficients of the transfer function denominator polynomial capture the persistence or decay of the effect over time
• A denominator polynomial with a value close to 1 indicates a strong persistence (the impact of the exogenous variable on the output series is long-lasting)
• The error term in the transfer function model represents the unexplained variation in the output series and is often modeled as an ARMA process
• The parameters of the error term, such as the autoregressive and moving average coefficients, provide insights into:
  • The structure of the residuals
  • The presence of serial correlation
  • The need for additional modeling of the error term
• Example: In a transfer function model for stock prices (output series) and economic indicators (input series), the coefficients of the numerator polynomial may indicate the sensitivity of stock prices to changes in the economic indicators, while the denominator polynomial may capture the persistence of the impact over time. The error term may be modeled as an ARMA process to account for any remaining autocorrelation in the residuals

Performance vs limitations of transfer function models

Evaluating model performance

• Transfer function models are evaluated based on their ability to:
  • Accurately capture the dynamic relationship between the input and output series
  • Generate reliable forecasts
• The goodness-of-fit of the model can be assessed using metrics such as:
  • Coefficient of determination (R-squared)
  • Mean squared error (MSE)
  • Root mean squared error (RMSE)
• The model's forecasting performance can be evaluated by comparing the predicted values with the actual values of the output series over a validation period
• Cross-validation techniques, such as rolling-origin or k-fold cross-validation, can be used to assess the model's robustness and generalization ability

Limitations and considerations

• Transfer function models assume a linear relationship between the input and output series, which may not always hold in practice
  • Nonlinear relationships may require alternative modeling approaches (threshold models, neural networks)
• The model's performance can be sensitive to the selection of exogenous variables and the specification of the transfer functions
  • Misspecification can lead to biased or inefficient estimates
• Transfer function models may not capture all the relevant factors affecting the output series, especially if there are unobserved or omitted variables
• The model's forecasting accuracy may deteriorate over longer horizons, as the relationship between the input and output series may change or be influenced by external factors not captured in the model
• Example: In a transfer function model for energy consumption (output series) and weather variables (input series), the linear assumption may not capture the potential nonlinear effects of extreme temperatures on energy demand. The model may also fail to account for other relevant factors, such as economic activity or population growth, leading to limitations in its forecasting accuracy