Dissertation
Statistical comparison of methods measuring context dependent functional connectivity
University of Iowa
Doctor of Philosophy (PhD), University of Iowa
Summer 2020
DOI: 10.17077/etd.005573
Abstract
As the field of cognitive neuroscience generates methods to answer questions about brain configuration during tasks, those methods need to be held to the fire to ensure they measure what they advertise and to help researchers understand under what circumstances the methods are appropriate to use. Without solid testing and validation of methods, the field could be led astray towards detrimental theories of brain function. I battle-tested several methods that measure context-dependent functional connectivity between brain areas. The methods were Beta Series Correlations (BSC) and Psychophysiological Interaction (PPI) analysis. In Aim 1 I compared two BSC variants---Least Squares All (LSA) and Least Squares Separate (LSS)---using both simulated and real data to ascertain which variant has more power to detect a change in context-dependent functional connectivity. While the simulations concluded LSS was the more powerful method, the real data painted a more nuanced picture. LSS appeared to have a disadvantage relative to LSA when the experimental contexts are interleaved with each other (as opposed to contexts being in separate blocks). LSS did not identify more false positives than expected when isolating brain areas highly involved in a task, whereas LSA did have a larger number of false positives than expected. I was also able to identify the optimal circumstances to use the BSC variants by simulating a variety of experimental designs. The simulations suggested at least 30 events per condition (with 45 being optimal), with each event being separated by an average of 6 seconds. In Aim 2, I compared BSC variants with a more established method named "Psychophysiological Interaction" (PPI) analysis to identify whether the BSC variants are more powerful than PPI analyses. PPI analysis was more powerful in the task I investigated. Additionally, the estimates from LSS are more correlated with PPI analysis than LSA, suggesting LSS and PPI analysis are more similar to each other. Overall, I found both BSC and PPI can likely detect differences in context-dependent functional connectivity and can be used with confidence when the experimental design meets recommendations.
Details
- Title: Subtitle
- Statistical comparison of methods measuring context dependent functional connectivity
- Creators
- James Kent
- Contributors
- Michelle Voss (Advisor)Eliot Hazeltine (Committee Member)Vincent Magnotta (Committee Member)Jatin Vaidya (Committee Member)Jan Wessel (Committee Member)
- Resource Type
- Dissertation
- Degree Awarded
- Doctor of Philosophy (PhD), University of Iowa
- Degree in
- Neuroscience
- Date degree season
- Summer 2020
- DOI
- 10.17077/etd.005573
- Publisher
- University of Iowa
- Number of pages
- xii, 156 pages
- Copyright
- Copyright 2020 James Kent
- Comment
- This thesis has been optimized for improved web viewing. If you require the original version, contact the University Archives at the University of Iowa: https://www.lib.uiowa.edu/sc/contact/
- Language
- English
- Description illustrations
- color illustrations
- Description bibliographic
- Includes bibliographical references (pages 138-156).
- Public Abstract (ETD)
We want to understand how the brain is organized, but we do not know how well we can trust the methods that help us answer that question. If we do not ensure our methods are accurate, we could lead researchers astray towards ideas about brain organization that are not useful. In this thesis, I create simplistic conditions where I know the truth to evaluate several methods that can measure how the brain is organized. I found under certain situations we can trust these methods which can help researchers move towards useful theories of how the brain is organized.
- Academic Unit
- Interdisciplinary Graduate Program in Neuroscience
- Record Identifier
- 9983987794202771
Metrics
82 File views/ downloads
77 Record Views