Off-campus UMass Amherst users: To download campus access dissertations, please use the following link to log into our proxy server with your UMass Amherst user name and password.
Non-UMass Amherst users: Please talk to your librarian about requesting this dissertation through interlibrary loan.
Dissertations that have an embargo placed on them will not be available to anyone until the embargo expires.
Author ORCID Identifier
N/A
AccessType
Open Access Dissertation
Document Type
dissertation
Degree Name
Doctor of Philosophy (PhD)
Degree Program
Organismic and Evolutionary Biology
Year Degree Awarded
2015
Month Degree Awarded
September
First Advisor
Benjamin H Letcher
Subject Categories
Biology | Statistical Models | Statistics and Probability
Abstract
Field studies that measure vital rates in context over extended time periods are a cornerstone of our understanding of population processes. These studies inform us about the relationship between biological process and environmental noise in an irreplaceable way. These data sets bring ``big data'' and ``big model'' challenges, which limit the application of standard software (e.g., \textbf{BUGS}). The environmental sensitivity of vital rates is also expected to exhibit interactions and non-linearity, which typically result in difficult model selection questions in large data sets. Finally, long-term ecological data sets often contain complex temporal structure. In commonly applied discrete-time models complex temporal structure forces the analyst to make many competing decisions about discretizing time, which add complexity to any analysis and confuse the interpretation of parameters. I tackle these problems in three ways: 1) I apply and improve current state of the art statistical algorithms and software (Hamiltonian Monte Carlo with the NUTS sampler implemented in \textbf{Stan}) to implement the estimation of vital rates from mark-recapture data in a large Atlantic salmon data set; 2) I extend continuous-time time-to-event models with continuous estimated rate functions to mark-recapture and emigration data; and 3) I apply simple local regression to the estimation of vital rates to turn the model-selection problem resulting from interactions and non-linearity into an estimation problem, which is solved by the estimation algorithm. For each model I present methods for visualizing the results in biologically meaningful terms and use these for model criticism.
DOI
https://doi.org/10.7275/7525804.0
Recommended Citation
Sakrejda, Krzysztof M., "Estimation Problems in Complex Field Studies with Deep Interactions: Time-to-Event and Local Regression Models for Environmental Effects on Vital Rates" (2015). Doctoral Dissertations. 510.
https://doi.org/10.7275/7525804.0
https://scholarworks.umass.edu/dissertations_2/510