[MUSIC] Hi, in this module I am going to be introducing respondent driven sampling. A sampling technique that is often used for the study of hidden populations, that may not be accessible via the traditional sampling techniques that we talked about in the previous modules. So for many populations, there are no systematic lists to be used as sampling frames. This may be because they are populations that no one is interesting in listing. Perhaps people with particular hobbies, fans of particular musical groups, players of particular video games. One way or the other, populations about whom there is little interest in compiling a systematic and exhaustive list. That we would normally consult in order to apply the random sampling techniques that we talked about in the previous modules. In other cases, and perhaps of more importance in terms of the usefulness for academic research, it is because the people are in the population in which we are interested in are deliberately hidden. Or rather they hide themselves, they want to remain anonymous. This may be the case if we're thinking about drug users, sex workers, other people engaged in illegal, illicit, stigmatized or perhaps embarrassing activities. Who try to remain anonymous when carrying out those activities. And for whom again, there is no exhaustive or systematic list that we could apply the traditional sampling techniques that we talked about in earlier modules too. So for such populations, these hidden populations one approach is respondent-driven sampling. The first step in respondent-driven sampling is to recruit an initial set of respondents whose serve as a sort of seed. This may be a convenient sample, where the researcher goes to a venue where known members of the population congregate. For example, if we are public health researchers trying to understand health risks associated with illegal drug use. And we are trying to develop an initial set of respondents, we might go to some venue, where drugs are sold. And try to recruit some respondents there, to serve as our initial respondents, our seed. These respondents are then asked to help recruit others in their social network, sometimes through the use of incentives. And then these new respondents, you might say the second wave, are asked to recruit new respondents who become the third wave. And this continues outward, to a fourth, or a fifth, or a sixth wave of respondents. Steadily further from the original seed respondents that we located perhaps, again, in a convenience sample at a venue. So let me give an example of how this might work. So if we start with again, hidden population that we are interested in studying, perhaps it is drug users. This is a common topic for public health researchers, because of the public heath implications of at least certain kinds of illegal drug use. So we might, as researchers, go to some venue where we know that drugs are bought and sold, and where drug users congregate. And at that location we recruit three respondents by approaching them, and asking to participate in the survey. When we have finished our survey, we then tell them that we are looking for additional respondents. And we may offer them a coupon, or a ticket, which can be cut in half, and which has code numbers printer on both halves. You tell them that they take these tickets and distribute them to other drug users in their network. And they retain half of the ticket with the code number, and then give the other half of the ticket to the member of their network. Then, if the member of that network comes in to see us bringing the ticket, we will match the code numbers on the two tickets anonymously later. And then they will both get incentives, that is the respondent who helped recruit the new respondent will receive some sort of incentive, perhaps a cash payment, as well as the respondent that they recruited. Again, through the anonymous matching of the code numbers on the tickets, so no names are involved. So, after we talked to these respondents, they go out talk to members of their social network, and they introduce our study to at least a few of them. Who come in with their halves of the ticket, and then we match the numbers, they receive an incentive, they complete the survey and so forth. And then we repeat the effort with them to recruit members of their social networks into our survey. So they introduce new people who come in, the tickets are matched again, incentives are distributed, and so forth. And this can continue until we build up, actually, a reasonably sized sample from that hidden population. So this is also known as snowball sampling. And theory and evidence actually suggests that after only a few iterations perhaps going out to a third, or a fourth, or a fifth wave. The resulting sample will be broadly representative of the targeted hidden population. This will be more successful in certain populations than in others. It will be more successful in hidden population whose members do interact with each other. So for example, sex workers that interact with each other, drug users that interact with each other, fans of particular musical groups that have contact with each other. This approach is more likely to work in such groups. It may not work as well in hidden populations whose members, in fact, do not interact with each other at all, because they prefer to be completely anonymous. So, again, it may not work in situations where the members of our hidden population are completely isolated from each other. So overall, respondent-driven sampling has emerged as an important technique in recent decades for sampling hidden populations. Either populations that are hidden because no one thought to enumerate or list them, and therefore, we don't have a traditional sampling frame. Or because the members of the population hide themselves and avoid as much as possible being listed. The technique is widely used now, especially in public health. People trying to understand drug use, sex work, and the implications for health of these and other behaviors. If you're trying to reach a hidden population, you should keep respondent-driven sampling in mind as an option.