New research from the University of Pennsylvania suggests that analyzing tweets on the social network can provide better insight into the prevalence of coronary heart disease in a community than many more traditional methods of prediction than factors such as smoking, diabetes and obesity — combined. The research was published last week in the journal Psychological Science.
University of Pennsylvania searched geo-tagged tweets send [sic] from 1,300 U.S. counties between 2009 and 2010, sorting tweets according to the types of emotions they conveyed. Researches then compared the findings to CDC heart disease mortality data from the same years. The tweets conveying negative feelings closely matched with the CDC data.
Tweets about things such as anger, stress and fatigue, it turned out, were a signifier for heart disease risk. More optimistic tweets, on the other hand, were associated with a lower risk of disease.
The article includes maps that show that regional patterns of CDC-reported mortality is well-predicted by the analysis of twitters.
Might it be that regional issues (poverty, political strife, crime levels etc.) ==> both heart problems and negative attitudes?
Or do regional issues ==> negative attitudes ==> heart problems?
Or do negative attitudes ==> regional issues ==> heart problems?
there is often this problem in deriving causal relationships from correlations.