There are a number of ways in which poor sampling could bias the results of a Nielsen ratings study. Let us look at some of them and at some possible steps that could be taken to minimize such bias.
First, poor sampling could lead to a study that does not accurately reflect all of the ethnic groups that watch TV. It could, for example, look at all groups equally, even if some are more likely to watch TV than others.
Second, poor sampling could lead to a study that does not look at people from all areas of the country. The survey might look too much at people from just one region. This would cause the study to be biased towards the sorts of things that people in that region chose to watch.
Third, poor sampling might result in a sample that was not representative of the various people of different ages who watch TV. Here, we might expect that older people will watch more TV. If a Nielsen survey does not include more older people in its sample, it might not accurately capture the viewing habits of Americans as a whole.
Fourth, poor sampling might simply bias the sample in some unknown way. For example, if Nielsen were to send out surveys by email, they would get a sample that is in some way self-selected because only those people who care enough to respond will be represented in the sample.
All of these potential problems can be solved by using proper, randomized samples of the correct population. Nielsen would need to carefully determine what percentages of people of different regions, ages, ethnicities, and other groups it wants in its survey. It would then need to select the proper numbers of people from each of these groups in a way that is random and unbiased. In this way, Nielsen would be more likely to get a proper survey with the minimum possible bias.