Predicting longitudinal post-traumatic stress disorder symptoms using psychological measures and heart rate variability: a machine-learning approach

Figures

• 

  Figure 1.. Scatter plots showing the means and standard deviations of the predicted versus actual PCL-5 scores across the 25 repetitions of the MKL regression models. (A) The upper panel (T1) illustrates the relationship between the predicted and actual PTSD symptom scores at the time of data collection, whereas the (B)lower panel shows the corresponding results for symptoms assessed 18 months later (T2). Each dot represents the mean predicted value for a given participant, and the vertical bars indicate the standard deviation across repetitions. The solid lines represent linear fits between the actual and predicted scores, thus illustrating the overall correspondence between the observed and modelled PTSD symptom severity.

• 

  Figure 2.. (A) The upper panel illustrates the kernel weights for the regression model at T1 and the (B) lower panel at T2. Each dot represents the kernel weight obtained in one of the twenty-five model repetitions with random fold assignments. The black horizontal lines indicate the means and standard deviation across repetitions.

• 

  Figure 3.. Item-level weights for the regression model at T1. The figure presents the weights of the top items from each of the four most influential psychological constructs. Each dot represents the kernel weight obtained in one of the twenty-five model repetitions with random fold assignments. The black horizontal lines indicate the means and standard deviations across repetitions. Note: Items marked with an asterisk (*) were reverse-coded prior to analysis, such that higher scores reflected greater levels of the construct (e.g. higher scores on ‘felt loved’ indicate greater emotional neglect).

• 

  Figure 4.. Item-level weights for the regression model at T2. The figure presents the weights of the top items from each of the four most influential psychological constructs. Each dot represents the kernel weight obtained in one of the twenty-five model repetitions with random fold assignments. The black horizontal lines indicate means and standard deviations across repetitions.