|
Do not necessarily confine yourself to one dependent variable. Often it is possible to think of different ways of getting at the same underlying construct. It may be feasible to measure some or all of these variables in your experiment and to analyse these variables either singly or in combination. That is, you can have more than one DV to assess the same effect variable. For instance, in the driving and music experiment (in Chapter 10 of the book) we might consider using a whole set of indices of driving performance - number of errors, mean speed, braking distance, and so on. However, if you do elect to use several DVs you must remember to address the problem of increases in your type I error rate with the number of separate analyses that you perform. (For more on type I error, see Section 11.3 of the book. See also Section F of this Web site.) There are statistics available to enable you to analyse these DVs simultaneously (multivariate analyses, such as MANOVA – See Section B7.4 of this Web site) or to assess whether it is appropriate to analyse a composite score (e.g., Cronbach’s alpha – See Section E9 of this Web site). |
