Fig. 6.

Distribution of time differences between the occurrence of the first post and the index date, for positive controls in the WEB-RADR reference set. Vertical bars indicate medians and diamonds indicate means. Note that only positive controls with at least one post have been included; the sample sizes are given for each dataset separately (with the total number of positive controls being 200). ‘Social 0.X’ means Twitter/Facebook data with a post-level indicator score threshold of 0.X. Forum posts were extracted with an indicator score threshold of 0.7. Data were collected from April 2012 and onwards