A Markov Regression Model for Nutritive Sucking Data

Abstract

We explore the relationship between nutritive sucking behavior and physiological status of premature infants. Conventional statistical methods are no longer adequate in this context due to the highly complex structure of nutritive sucking data. We describe sucking pattern by a binary time series and propose a Markov regression model that relates the transition probabilities of the Markov chain to time-dependent covariates through a logistic linkage function. The model is fit to data collected at the Hospital of the University of Pennsylvania. The results are compared with those obtained from traditional data analysis.