Available at: https://digitalcommons.calpoly.edu/theses/3043
Date of Award
6-2025
Degree Name
MS in Statistics
Department/Program
Statistics
College
College of Science and Mathematics
Advisor
Kevin James Ross
Advisor Department
Statistics
Advisor College
College of Science and Mathematics
Abstract
Understanding the interaction between mother and baby during feeding is critical for the long-term development health of the baby. Overfeeding can lead to later obesity, while underfeeding can lead to malnutrition. In a recent study, the behaviors exhibited by mother-infant dyads across multiple ages of infants have been observed and coded according to the Baby Behaviors when Satiated (BABES) coding scheme. However, creating models using the data obtained from this coding is no simple task since the data coding is continuous, multivariate, and longitudinal in nature. The specific model utilized for these data is a hidden Markov model, since there are multiple behaviors for the mother-infant dyads that are either happening or not at a given time. Identifying hidden states in the model enables a form of dimension reduction; rather than enumerating all possible combinations of behaviors as observed states, analysis can instead focus on the smaller set of identifiable hidden states. Hidden Markov models were fit to the data at six different ages by using the mHMMbayes package. It utilizes Bayesian estimation and accounts for multilevel framework and categorical multivariate data. Using these model results, hidden states can be described by viewing the posterior transition matrices and emission distributions. These model results will help understand the feeding behavior in mother-infant dyads on a deeper level.