Revolutionary wrote:Okay, so my faggot doctor wants me write about ways to improve the way we collect data and the way we choose our target audience as well as how to increase the level of accuracy and honesty amongst the population when we conduct a research.
This is pure medical talk, but I guess "research" is research no matter what we're talking about.
Any ideas?
Improve the ways we choose our target audience
It all depends on what conclusions you want to make.
So, if you want to be able to make very general conclusions about the whole population, the best way to is probably "simple random sampling", and a relatively large sample size, so your sample isn't biased. It's important that it is random. Theoretically, everyone should've had the same probability of being selected.
Or, if you’re interested in a specific subpopulation, with some attribute X. Then you might want to consider a sampling strategy that focuses on these Y people with attribute X. You could, for example, conduct a pre-research in public databases in order to identify people Y with attribute X.
How to increase the level of accuracy and honesty
… Off the top of my head:
1. Stress the fact that your answer is very important to this research, and future research. “You are unique, and your answers cannot be replaced”.
2. Stress the fact that you can have an influence on X (whatever X is), so that X is improved according to your preferences.
3. Stress the fact that the information will be treated with respect and anonymity.
4. Stress the fact that only summarized information will be published. In other words, a random reader will not be able to identify that person X answered Y on question Z.
5. Create an incentive for person X to answer survey Y. It could be some form of reward or the chance of a reward… It could be as simple as the "chance of winning" cinema tickets.
6. Have a serious and professional attitude, if it's physical survey, you should put an emphasize on design.
7. The way you construct questions, should be in an objective manner. I.e. you shouldn't form questions so that the answering person is more likely to answer X than Y on question Z.
8. Think about the “scales” for your questions. For example, you might have a question: “What do you think of X?” Then a reasonable scale could be 1-5, with 1 = Very bad, and 5 = Very good. If you instead would’ve gone with 1=Bad, and 5=Very Good, you would have implicitly created a bias towards positive responses – in other words, decreased the accuracy.
9. Keep the amount of questions relatively small; not more than 20. People lose interest fast…
10. Keep the questions simple, and understandable. This is very important for accuracy. Because, if you are able to reduce the probability of people misunderstanding your questions, you’ll also reduce the risk of people answering “incorrectly”, and consequently also increase the overall accuracy of the whole survey. Also, if people think the questions are too complicated, there’s also a risk that they’ll just put the survey away, and don’t answer it at all.
11. Conduct a post-research study, analyzing data, and dismissing observations (people) with unreasonable answers. This is quite easily done with residual analysis.
