Assessing the bias in samples of large online networks