In linear regression, the observations (dots) are assumed to be the result of random deviations (strings) from an underlying relationship (stick). A simple but highly visual practical implementation of the "line of best fit" [more: https://t.co/ASnVxNMevD]