A newspaper wants to gauge public opinion about legalization of marijuana. The sample size formula indicates that it need a random sample of 875 people to get the desired margin of error. But surveys cost money, and it can only afford to randomly sample 100 people. Here’s a tempting alternative: If it places a question about that issue on its website, it will get more than 1000 responses within a day at little cost. Is it better off with the random sample of 100 responses or the website volunteer sample of more than 1000 responses?