Only if you have reason to suspect that those who opt-in are a skewed sample with respect to the things you want to measure.