A: We've primarily made use of 3 methods: school-based samples, random-digit dialing, and list-assisted samples. There are pros and cons to each.

A: Well, the school gives you a list of names, so it's a simple and fairly inexpensive method to use. And, if your interest is in kids who are in school, the lists are usually complete. But it's important to realize that you're also excluding kids who don't go to school, drop-outs, truants, kids who've finished school, and kids in alternative schools. Since we are studying high risk behaviors like drinking, if we relied only on school-based lists, we'd be missing some of the kids we want to target.

Another disadvantage is that it's getting harder to survey students on risky behaviors because the school districts don't allow certain sensitive topics to be surveyed at all - some districts won't even provide students' names and addresses for research outside of school. All of this makes it more difficult to do school-based survey sampling, and may even bias your sample, if these district-level decisions aren't random.

A: In RDD, you start with a randomly-generated list of working phone numbers. When you reach a potential participant, you determine if there's a child in the household in your target age group, and then you go from there. The advantages to this method are that you don't have to depend on the schools, and that it's more comprehensive ñ you have the chance to include drop-outs and others that wouldn't be in the school lists.

A: It can be quite costly and inefficient. In one of our studies, we had to call about 12 households to get one potential participant. And people are less likely to tolerate telephone surveys these days, not to mention caller ID and caller screening, which make it much harder to make contact. There's also a bias to RDD for two reasons: you don't know the characteristics of those you weren't able to contact, so there's unknown bias there, and you may get fewer low SES and minority participants using RDD because they may not have working phones.

A: No, the new telemarketing law does not affect RDD surveys if they are for research and not commercial purposes.

A: Estimating your response rates is difficult, since you don't know how many non-contacted households were eligible. You will greatly overestimate response rate if you divide the number of completed interviews with the number of contacted, eligible households. So, you need to approximate how many of the households that you couldn't reach would have been eligible. This probably underestimates response rates a bit, but usually you just assume that it's the same proportion of eligible/non-eligible households as in the contacted group.

A: You buy a list of phone numbers that have a higher probability of being eligible households. They're often pre-screened to eliminate nonresidential numbers, and are based on consumer information like magazine subscriptions. Oftentimes you can also obtain address information. It's more efficient to use these lists than to do RDD, obviously, since you have a better chance of reaching your target population. The disadvantages are, like RDD, that you may not get a representative sample, and that it's hard, again, to estimate response rate. But a big advantage in this method is that you can send letters in advance to the households telling them about the study and inviting them to participate. This can lead to higher cooperation rates than if you just cold-call households.