Threshold Autoregressive Models Show Significant Small Sample Bias in Real Exchange Rates.
The study looked at a type of model called Threshold Autoregressive (TAR) models that deal with nonlinear dynamics. They found that when studying exchange rates, these models can have a significant bias when using small amounts of data, even if the data covers a long period of time. This means that the results from these models may not accurately reflect the true relationships between variables.