1

I have a set of Likert type items asking respondents their attitude on different topics. For example, attitudes toward drugs/alcohol where 1=strongl disagree and 5=strongly agree to the statement "I think it is okay for someone my age to smoke marijuana." The 6 items in this particular scale are then scored using the mean of non-missing items. Is there some advantage to first rescaling the items so that the range of responses goes from 0-4 rather than 1-5? The only difference is in interpretation of the scale where the average "attitude" is 3.5 on a scale of 0-4 and 4.5 on a scale of 1-5. From an interpretation standpoint, what is easier for a lay audience? Or is there some standard of practice in terms of reporting?

PCharles
  • 11

1 Answers1

2

I would think that a scale of 1 to 5 would be a more advantageous scale since zero is a non-intuitive number. Zero has its own connotations such as representing nothing or an absence of the value being measured. From your example, a scale of 0 to 5 only "sounds" intuitive to me if 0 corresponds to a value which represents "never".

In addition, whenever you overhear someone say one a scale of ... to ... it is almost always a scale of 1 to 10 or 1 to 5. I would stick with 1 to 5.

Francis Smart
  • 610
  • Thanks much. I totally agree but have seen more than once colleagues who prefer to rescale the scores despite the scale not having a true zero or the lower bound score representing "never". These scales are all just agree-disagree scales. Appreciate your comment. – PCharles Mar 22 '14 at 21:55
  • I once set up a scale: -2 Strongly Disagree, -1 Disagree, 0 Neutral, 1 Agree, 2 Strongly Agree. That seemed more appealing to me since it was easier to grab statistics such as what percentage were negative/positive on a question. It also creates a kind of internal consistency in which the sum is always positive if more people strongly agreed than strongly disagreed. – Francis Smart Mar 24 '14 at 00:40