The second, and very different, new data source that we saw over the last few years, are these Access Panels. New in the sense that in the online format, they're really spreading widely. And, this link here on the left hand of the slides, gives you links to a lot of different surveys here. And we have just a screen shot of a long list. Colleagues of ours working at Market Systems Group, for example. But, many, many, many others as well. They have a set of people, and actually so does SurveyMonkey, our partner in the specialization. They have pre-recruited people, constantly recruiting people in many different formats. This can be deliberate recruitment, could be even a probability of recruitment for some of these panels. But also, pop-up menus when you visit a website, inviting you to join a panel, or advertisement of other sorts. Good panels often try a variety of recruitment strategies so that the set of people, and the composition of the people, is quite diverse. The advantage of these Access Panels is, they have a large pool of people, and for every given data analysis, for every particular survey you wanna do, there's a smaller group that's taken out of it and the survey is fielded to. Usually, that then is pretty cheap, but keep this format in mind as we talk in module three about the different possible error sources. So that you then, for your own research question can decide, will this be a problem here or not? If you think back to module one, when we talked about description, was this causation, was this prediction? These Access Panels might not be ideal to correctly give you a proportion of people who suffer from disease x. Or in particular, if something is related to literacy they might not have anyone in there at all. Because those people do not participate in Access Panels if the delivery mode has all to do with reading and writing. Same with age distributions and the like. Now, for most of these demographic groups, they actually are not that bad. The more common these panels get, or depending on your research question, every single time you will need to ask yourself before is, the selection mechanism with which people participate in these panels, related to what I want to do? My area of interest and is there a chance that it can be distorted? Now, as I said, if you are, for example interested in doing a causal analysis. And you would use the Access Panel to field an experiment, let's say a question of wording experiment. Then unless you think that the self-selection into this Access Panel would distort the wording experiment that you do, you're probably fine with your results. So that's why module one, we made sure you understand the different purposes of your data collection. Now we introduce several different modes to you so that you have a platform up on which you can think about the different [INAUDIBLE] we talk about in module three. There are a variety of Access panels, not just for market research but also a set of academic access panels. The list panel in the Netherlands is one example, it's run by CentERdata. It is based on a probability. Respondents without internet access actually did get a computer or some device with which they're able to access the internet, and can participate in the panel. This much more mimics a regular survey, with the sampling procedure. And so here too, knowing as much as possible about the data generating process. Transparency of the process will be key so that you can evaluate quality of the data. If you wanna read up on this Academic Access Panels or another Access Panel and quality of studies that assess their quality that compare those panels with other types of data. You can use this resource, WebSM. It's a fantastic platform that has almost everything that has been published to this point either referenced here or with a direct link. Coming now to the Google Consumer Survey, another very different type of data collection here. It's very few questions spread out to the respondents. If you look at the Google Consumer Survey, you want to field one yourself. You'll see you can select demographic characteristics. Many of these things, sort of guesstimates from profiles. So, interesting methods report on the Google website on this. And then you can create your own survey as you can see here on that platform. Many more survey creation features are available, for example, on websites that are particularly designed to do that. Like SurveyMonkey or Qualtrics. LimeSurvey's another one. So, there's many providers of these, these days. If you plan to use a web survey, it's certainly not a bad idea to start with those rather than trying to program everything from scratch. It can take a lot of resources, and it might be cheaper to use these. So, if you actually plan to do a web survey, it's of importance that you keep general principles in mind here, how to Design Effective Web Surveys, like the book published by Mick Cooper that you see here on the left. Or other summary of the scientific findings that Roger Tourangeau, Fred Conrad, and Mick Cooper put out. Or the just recently published book on Web Survey Methodology by Mario Callegaro and his colleagues, Katia and Vanja, that the later two actually boast being responsible for webMS that I showed to you before.
