Close Menu

Using Relationships from Misaligned Data to Improve Predictions

Using Relationships from Misaligned Data to Improve Predictions

Dr. Kelly-Ann Dixon Hamil
Faculty of Social Sciences
Economics
Theme: 
Finance and Logistics

Introduction

In order to maintain privacy, social, economic and health data are published based on geographic boundaries. However, agencies collect and report these spatial data using the boundaries and/or areas that benefit their mandate (see Figure 1). This results in the misalignment of boundaries when data from different agencies are used to model relationships. Despite this, for effective policy and proper planning, this disparity must be accounted for in any such analysis.

Methodology

The method uses a joint semiparametric (parametric + nonparametric methods) to modelthe processes (i.e. variables) of interest(see Figure 2).

Results

This method improves predictive performance when compared to using other methods.

Conclusion

The semiparametric method:

  • is effective in improving the predictive performance when examining two spatial variables that do not have a hierarchical structure
  • provides a solution for common challenges experienced when using social and economic spatial data
  • is flexible and can be used with different spatial correlation structures.

Relevance and Potential Application

This method increases the ability of researchers to examine relationship between variables which usually require the reconciliation of misaligned boundaries. Some examples include: the relationship between voting patterns and expenditure on education or health services; the incidence of cancer and temperature and many others.

Top of Page