Testing for omitted variables in a diffusion model with an application to term structure of interest rates

Payments Canada recently released its latest discussion paper, Testing for omitted variables in the diffusion matrix of a multivariate diffusion process with applications to term structure of interest rates. This paper provides tools that can be used to validate time-series regressions used in nowcasting.

This research has been developed to detect the potential misspecification of the volatility in a high-dimensional time-series model that could result in spurious forecasts when using such models. High dimensional time-series models are typically used in predicting trends in macroeconomic variables like GDP, inflation, and unemployment as well as the relationship between stock prices and volume.