The survey is always offered only to a random subset of Steam users. The results only ever represent the fractions of users who took the survey, and are not representative of the entire Steam ecosystem as a whole. Unfortunately, this means that the increase/decrease in Linux usage is probably within the margin of error and is not a reliable statistic.