Analysis code and source data for Spatial Navigation as a Digital Marker for Clinically Differentiating Cognitive Impairment Severity.

This repository contains the sanitised analysis code and source data underlying Figures 2 and 4 of the manuscript "Spatial Navigation as a Digital Marker for Clinically Differentiating Cognitive Impairment Severity".Code files ANOVA&suppl_info.qmd – Quarto document reproducing Robust ANOVA across CDR groups associated with Figure 2 and supplementary statistical analysis including Spearman correlation matrices (Good vs. Poor path integration groups) with heatmap visualisations and robust regression to assess the effect of tablet experience on performance in SPACE.AUCs.qmd – Quarto document reproducing statistical analyses across CDR groups, including robust ANOVA, AUC estimation across CDR levels, models with and without path integration, comparison of SPACE against neuropsychological tests, forest plot generation, and clinical diagnostic performance analyses (AUC, sensitivity, specificity, Bonferroni correction).clinical_characterisation.r – R script producing the clinical characterisation summary table, including descriptive statistics and between-group comparisons across CDR categories for demographic and clinical measures.crossvalidation.R – R script implementing cross-validation procedures for logistic regression classification models used to evaluate diagnostic performance across CDR group comparisons.The analyses include:Robust ANOVA across CDR groupsLogistic regression classification modelsArea Under the Curve (AUC) estimation with two-sided 95% confidence intervals using the non-parametric DeLong methodGeneration of boxplots and forest plots reported in the manuscriptSource data files Figure2_Source_Data.xlsx – Numerical values underlying the boxplots shown in Figure 2, including group-level summary statistics for training time, path integration distance, and perspective-taking error.Figure4_Source_Data.xlsx – Numerical values underlying the forest plots shown in Figure 4, including AUC estimates and corresponding two-sided 95% confidence intervals for all pairwise CDR group comparisons.