Quote:
Originally posted by Thoran:
Well I'm sure there's volumes on how to get a decent control for statistical studies... but in this case I'd suggest mailing surveys to random households in a region covering roughly the same area as the respondents. If you solicit online for a control, you'll be skewing your control group based on the fact that they're online... to obtain a truly population representational sample you need to go snail mail. Even then you'll have to discuss in your results the several problems inherent in gathering data by these methods and any techniques employed to mitigate those issues.
|
Consider though that his experimental group are online - their results may be different to gamers who are not online. It might be more appropriate to have a comparison group that is online - either both are online, or neither. You have to make sure in a study like this that your control group and experimental group are equal in every way except the manipulation, in this case whether they're gamers or not. Otherwise your results are open to confounding from other variables - that is, you can't draw any conclusions because effects might be due to other variables not the one you have manipulated. There may be something about being online that affects your daydreaming that not being online does not. A good idea if you had the time would be to get a group of gamers offline and a group of non-gamers offline (or use the control data) and compare to see if there are substantial differences between your online and offline samples. If there are, then your results for the online sample aren't generalisable to the population.
I agree though that you'll need to consider what getting data online like you have means for your results. While it's unlikely that people would lie about their age and so on simply because they're online, it is a fundamentally different way of finding participants than, for instance, stopping people in shopping centres where you can see them. It's also important to whether your sample is representative of the population. If they are random and aren't systematically different from the population, then results from them are generalisable. If being online means they will be systematically different from non-online people, then you can only say your results are valid for people who are online.
Your normalised t-scores would simply the scores of a random sample from the population that the test was normed on, that is, tested on to see what the 'average' responses were and what the normal population of scores is. In the population, there will be people who score high and score low, with most people scoring around the average. The idea of your study then, is to see whether the mean for gamers is significantly higher or lower than the mean for the population.