Computational protocol: Quantifying the impact on navigation performance in visually impaired: Auditory information loss versus information gain enabled through electronic travel aids

Similar protocols

Protocol publication

[…] To assess differences between the modalities (‘Normal’ vs. ‘Auditory Information Loss’ and ‘Normal’ vs. ‘ETA’) for each of the four endpoints, we used linear mixed models (LMM) with subjects as random effects due to the repeated measurement design of the study. Even though we were only interested in differences between the respective modalities, the model included both modality and variant as fixed effects because the maze variant also constitutes a known source of variation. We also tested for interaction between these two factors but found that this term could be dropped from the model. For contacts, a Poisson model was also considered but is not presented here since it yielded similar results as the linear model. Data are presented as means for the reference modality ‘Normal’ and mean differences for the other modalities (each with 95% confidence intervals), as well as with boxplots.In addition, we investigated the distribution of relative viewing angles of each subject to show individual head movement characteristics. A narrow distribution would indicate that the subject gazed mostly in the direction of the intended walking path whereas a broad distribution would be characteristic of subjects with more expansive head movements. Distributions with a high number of large angles represent subjects who frequently experience orientation losses. As an exploratory analysis it was thus investigated whether subjects whose passage times improved with ‘ETA’ showed different viewing patterns than subjects whose passage times were slower with ‘ETA’ (as compared to ‘Normal’). This grouping was achieved by comparing individual subjects’ median passage times for the two modalities. For the exploratory analysis, we used a mixed model for the endpoint MARVA considering only runs with modality ‘Normal’ or ‘ETA’. The model included all fixed and random effects from the main analysis but also contained group and an interaction term between group and modality as additional fixed effects. The same analysis was repeated with the endpoint percentage of orientation loss to find out whether differences are based on actually different viewing behaviors or rather on an increased number of orientation losses. Finally, to further analyze the differences between groups, we investigated the correlations between passage times and number of contacts within each modality (‘Normal’ or ‘ETA’) and group. Individual runs were pooled by their median and Spearman correlation coefficients were calculated. Mixed model analyses were performed using R version 3.2.2 (packages nlme and lme4) and SAS version 9.4 (proc MIXED). A p value of <0.025 was considered statistically significant due to separate tests for the modalities ‘Auditory Information Loss’ and ‘ETA’ respectively. […]

Pipeline specifications

Software tools nlme, lme4
Application Mathematical modeling