SC @ 2008-04-10:
Could I ask you a question? I measure a variable of xxx twice in one study - time 1 and time 2, but using different scales, for the purpose of preventing the subjects from guessing or lazily using the time 1 score. In time 1, I use a 10-point Likert scale (1 strongly agree -- 10strongly disagree); in time 2, I use semantic differential scale (1 very dissatisfied -- 10 very satisfied). The DV is the result of time 2 minus time 1.
Do you think using such different scales is reasonable in experiments? Or should I use same type of scale both in time 1 and time 2?
庄主 @ 2008-04-11:
Your dilemma is between instrument effects (i.e., carry over effects of test1) and instrument validity (i.e., equivalence between Likert scale and semantic differential). In general, the former is considered to be less an evil because the possible presence of carry over effects would lead to support your null hypothesis (Type II Error) whereas the use of two fundamentally different instruments would lead to reject your null hypothesis (Type I Error). >>全文