Abstract
dc:description.abstractWe often want to achieve our goals as effectively as possible. But what are the most effective interventions to achieve our goals? Here, we often turn to science for advice. And, scientists across various fields – from social science to medicine – measure how effective tested interventions are, using so-called effect size measures. Such effect sizes are widely used to inform evidence-based decision-making. But do the effect sizes of tested interventions convey the information relevant for rationally choosing the best intervention? The arguments I present in this thesis suggest otherwise: Effect sizes do not always convey the information that matters to rationally choose between interventions, and effect size measures differ in when they provide the information relevant for rational decision-making. Based on my results, I advance debates on which, if any, effect sizes scientists ought to use to inform decision-making. In Chapters 2 and 3 I focus on effect size measures for binary outcome variables. In Chapter 2, I argue that so-called absolute effect sizes are at least as good as, if not better than, so-called relative effect sizes for informing rational decisions across choice scenarios, at least for some rational agents. But, as I show in Chapter 3, both absolute and relative effect sizes may still omit information that risk-sensitive rational agents care about in deciding between interventions. These findings advance recent debates on whether medical researchers should seize to report only relative effect sizes. In Chapters 4 and 5 I focus on effect size measures for continuous outcome variables. In Chapter 4, I show that even perfectly accurate so-called mean differences omit information which can be vital for people to rationally choose the best intervention. However, the good news is: I provide sufficient conditions for when mean differences do not omit information that matters to rational decision-makers. In Chapter 5, I focus on the widespread practice of using so-called standardised mean differences to measure an intervention’s effect size when researchers face measurement uncertainty. I argue that reporting standardised mean differences risks omitting information that matters for rational decision-makers, information they could learn from the mean differences and standard deviations researchers could instead report. These findings advance longstanding debates on how researchers in the human sciences should use mean differences and standardised mean differences to inform evidence-based decision-making.
Degree
thesis:*- Name dc:type.qualificationname
- Doctor of Philosophy (PhD)
- Level dc:type.qualificationlevel
- Doctoral
- Grantor dc:publisher.institution
- University of Cambridge
- Year dc:date.issued
- 2024
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Jäntgen, Ina
- Advisors dc:contributor.advisor
-
- Bird, Alexander
- Stegenga, Jacob
Subjects
dc:subject × 4Rights
dc:rightsIdentifiers
dc:identifier.*- DOI dc:identifier.doi
- https://doi.org/10.17863/CAM.115922
- OAI identifier oai:identifier
- oai:www.repository.cam.ac.uk:1810/379952