# Sampling frame of a voluntary survey

If one is conducting a survey in which all the respondents are volunteers, would there be a true sampling frame?

I see three possibilities:

1. We could view this as a situation where there is no set list of subjects from which those that end up being included in the sample are chosen -- so therefore there is no sampling frame

2. The sampling frame could be every possible individual in the world that has access to a computer (it it's online) or exits in physical space and who might encounter the person administering the survey- say anyone in the world who might live in or might be visiting the city and might happen to walk by the corner where the person who is administering the survey might be standing. This would just mean that the vast majority of the sampling frame would have an extremely low inclusion probability

3. The sampling frame is all people who have a reasonable opportunity to take the survey- so if it were an internet survey, all of those who were asked to take the survey by email or had seen a page that had a link to the survey. Similarly if someone were administering a survey in say a mall, then the sampling frame would be all people who were shopping in that mall in a given time.

Your first possibility is correct-there is no frame. Your second and third define potential populations of volunteers. A frame is not a population-rather it is a method that allows you to link random numbers to (ideally) every member of the population.

Then a draw of random numbers (perhaps in several stages) will identify the sample. Many people think that a frame must be a master list of the population members, but that's often not the case. A multi-stage area sample, for example, starts with a list of areas (primary sampling units or PSUs), and random numbers draw a sample of these. In each selected PSU a list of smaller areas is prepared, and a random sample of these is taken. The process can continue down to the selection of households and of individuals in households. The process is complicated when populations shift over time-consider post-voting surveys. In such cases, the method is to take a systematic sample of voters in time order.