Yang: Even attempting to determine the percentage of error in a non-probability sample (self-selected) is futile.

No disagreement. When I mention a sensitivity test, even that would be meaningless at this rate of response. Example: For a simple one-question survey with 1500 responding out of possible 1600 and a 60% (900/1500) "Yes" vote, you could assert that the true "Yes" is somewhere between 56 and 62 percent (b/w 900/1600 and 1000/1600). That's potentially useful information. If it were a 60% "yes" within 110 (out of 1600 potential) responses, well, I don't think there's any useful infomation there (except that you have a horrible response rate!!)