Set of basic statistics for direct implementation and visualization

To obtain the basic statistics for visualization in WP2 and implemented in WP7 as well as to provide a solid baseline to which to compare the more advanced methods below WP4 will first focus on simple quick to compute statistically robust low standard error and easy to interpret statistics that describe distributional properties of the data and covariation between different feature dimensions These involve mean levels and variability of symptomscontextual factors over time as indicators of symptomcontext level and volatility and their mutual comparison to detect key personalized strengths and weaknesses Next they involve simple statistical time series tools for tracking symptoms and contextual factors and their correlations over time such as autoregressive movingaverage ARMA models or measures of mutual predictability Granger causality allowing to obtain basic insight into how symptomscontext variables cooccur or predict one another over time Finally to capture clinically significant moments of change or tipping points within the behavioural trajectories as they are unfolding across larger periods of time WP4 will use statistical change point detection techniques as developed in Leuven eg 138 and Mannheim eg 139 to detect reliable phase changes in mean levels as a function of for instance instalment of treatment or a particular change in treatment over time Methods for the correction of familywise error rate such as the HolmBonferroni procedure regularization techniques in model estimation and measures of outofsample prediction error will be adopted to minimize the risk of false positives to inform clinical decision making Outcomes from lowerlevel statistics and machine learning predictors will be summarized and delivered to the visualization platform in a clinically meaningful and accessible way