Designing a psychological test isn't extraordinarily difficult, as long as you have access to a hundred or so people who have the trait you're testing for and another hundred who don't.
- Figure out what you want to test for. See what the symptoms are and find out what its related to. Before you start you want to have some evidence that the trait you're testing for isn't a figment of your imagination.
- Make up a long list of questions--at least 3 or 4 times as many as you want to test to include at the end. Make sure the questions don't have excessive "face validity" (meaning, don't make them too obvious!) If the questions are obvious, the results will reflect self-image, or the image the test taker wants to project, and so it won't necessarily be accurate.
- Give the questions to both groups and compare the results. This will take some statistical analysis, but all the necessary tools are available in Microsoft Excel and, anyway, don't require any math more complex then high-school algebra.
- If a question is answered in significantly different ways by the two groups, keep it. If not, then drop it. Also, check the questions across groups for factors--see if the same people tend to answer a group of questions the same way. That indicates that those questions are picking out a particular trait distinct from the one you're after. This is a little tricky, because a "factor" in your test could be one of several things: an irrelevant confounding factor, another personality trait that's related, but not identical, to what you're studying, or a subset of your test.
- Take your final draft of the test and give it to another group of people. Look at the distribution of scores to figure out a scale that usefully identifies the trait you want to test. If you found useful factors, you might create a main score and a couple of sub-scores.
- Now get your test published--that can be the hard part.
See also
the inherent limitations of psychological testing.