Far East Journal of Theoretical Statistics
Volume 14, Issue 1, Pages 57 - 77
(September 2004)
|
|
IMPROVING THE PERFORMANCE OF A STOCHASTIC MODEL BY INCORPORATING A NON-LINEAR TERM: APPLICATION TO A SPIKING BIOLOGICAL SYSTEM
V. K. Kotti (Greece) and A. G. Rigas (Greece)
|
Abstract: In this work we study the behavior of a complex biological system called muscle spindle that plays an important role in the initiation of movement and the maintenance of posture. Both the input and the output of the system are binary time series and therefore the system can be identified by using Generalized Linear Models (GLMs). Three functions of special interest, the threshold, the recovery and the summation are estimated by the maximum likelihood approach. Two cases are considered: (a) a constant threshold model and (b) an improved model with an exponentially decaying threshold function. These models are applied on two different experimental situations. The validity of the two models is checked and their performance is compared by the means of deviance. |
Keywords and phrases: binary time series, deviance, generalized linear models, linearization technique. |
|
Number of Downloads: 5 | Number of Views: 732 |
|