Adaptive Multivariate Autoregressive Modelling of Respiratory Cycle Variables

In 1963, Priban1 showed that the breath-to-breath variability in the respiratory cycle variables is not purely random. We were interested in studying this variability using autoregressive models as Benchetrit and Bertrand2 did, but we found that often the data series were not stationary and therefore could not be analyzed using traditional autoregressive (AR) modelling techniques. What will be presented here is a method that has been developed for fitting AR models to non-stationary data by fitting the models adaptively.