Using linear mixed models to analyze data from eye-tracking research on subtitling

Abstract

Abstract In this paper, we aim to promote the use of linear mixed models (LMMs) in eye-tracking research on subtitling. Using eye tracking to study viewers’ reading of subtitles often warrants controlling for many confounding variables. However, even assuming that these variables are known to researchers, such control may not be possible or desired. Traditional statistical methods such ast-tests or ANOVAs exacerbate the problem due to the use of aggregated data: each participant has one data point per dependent variable. As a solution, we propose the use of LMMs, which are better suited to account for a number of subtitle and participant characteristics, thus explaining more variance. We introduce essential theoretical aspects of LMMs and highlight some of their advantages over traditional statistical methods. To illustrate our point, we compare two analyses of the same dataset: one using at-test; another using LMMs.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 26 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.