It'd bad to rely on a small selection-biased sample of anything as an indicator of the general case, regardless of whether the sample data is acquired firsthand or secondhand.