Fatima Gallardo Ibarra- 2023
Files
Publication Date
11-10-2023
Digital Publisher
Digital Commons at St. Mary's University
Disciplines
Higher Education
Keywords
T-test; independent-sample t-test; paired-sample t-test; missing data
Creative Commons License
This work is licensed under a Creative Commons Attribution-NonCommercial-No Derivative Works 4.0 International License.
Collection
McNair Scholars Symposium
Format
MOV
Medium
video
Description
Missingness interferes with practitioners’ ability to accurately interpret the results of their studies. This is especially true in within-subjects designs as a study with a sufficiently large original sample size might net a much smaller set of complete cases once missingness is taken into account. Contemporary solutions to dealing with missingness, such as regression and multiple imputation, have focused on replacing missing values. Methods such as these that replace missing values rather than delete cases altogether preserve sample size, but they also require a level of sophistication that far exceeds what practitioners can reasonably be expected to have. Recent research has supported the use of between-subjects methods in within-subject scenarios as a means by which to manage missingness in a highly accessible manner. The present study builds on this earlier research by considering whether a new strategy, Optimal t, maintains statistical power in the face of missingness without artificially inflating alpha. Optimal t selects the better option between a paired samples and an independent samples t-test based on simulated trials. Initial research has supported the use of Optimal t, although there has been some concern about inflated Type I error rates (Sperling et al., 2023). In this study, a correction formula is introduced which adjusts p-values in proportion to sample size and compared the results against original Optimal t, paired samples t-test and independent samples t-test. Results suggest that original Optimal t outperforms all other options across conditions. Implications for practice and recommendations for further research are provided.
Loading...