A new class of Smooth Transition Autoregressive models, based on cubic spline type transition functions, has been introduced and subjected to comparison with models based on the traditional logistic transition functions. A very high degree of similarity between the two model classes has been demonstrated.
The new class of models can be slightly preferable because of its more simple formal and geometrical structure that may enable users more convenient manipulation in statistical inference procedures.
Keywords: time series; regime-switching autoregressive models; logistic and cubic-spline transition functions;
AMS: 62M10;
BACK to VOLUME 40 NO.1