Regards. In the analysis of data from an attitude scale (Likert) what level of measurement must be selected in view of variables: or ordinal scale? My question is because some authors have selected in the view of variables as scale (no tag value) and other authors as Ordinal. There are also those who considered the ordinal variables (dimensions, indicators or items), but the sum or scale transformation consider it? Thanks in advance.
Good question. Technically speaking, most if not all Likert scales are ordinal and the most conservative approach is to use nonparametric statistics. The disadvantage is that there few multivariate nonparametric tests, which precludes the use of advanced analyses, e.g., MANOVA, regression models, et al. When parametric and nonparametric analyses are applied to the same data, e.g., Likert scale data, the results typically yield the same conclusion. Because of this, most researchers use parametric statistics to analyze these data. Journal editors recognize this and accept papers that use them.
Cheers and Thanks, Donald
Sorry to insist. And the analysis of data from an attitude scale (Likert) with SPSS, please tell me what option should I select: or Ordinal Scale? Thank you
Ok. I see. Select "ordinal." Compare results to those when you select higher level measurement. If results differ, stick to ordinal. If not it won't matter
I never thought it mattered much which you use, until a colleague showed me how, w/ certain datasets the conclusion about the null hypothesis would be different w/ parametric vs. nonparametric stats! Yes, that's right, you could make the wrong conclusion w/ the same dataset by using the wrong statistic.
Since then, I have tried to be more cautious. However, if a colleague wants to do parametrics on Likert for the reasons the first person answering mentioned, I don't resist too much. I do prefer to see what the nonpar would have said also though. That also means I am ready in case any reviewer wants to bring up that we should have done nonpar.
My 2 cents.
Technically speaking, Likert-scale items are ordinal because there is not equal distance between intervals. In conservative or hard sciences, ordinal would be necessary. However, in behavioral and social sciences, Likert-scale items are regarded as interval level items (scale). In these sciences, the assumption is that the participants are thinking in a similar fashion about the distance of intervals when responding if the item is framed well. Although this can be problematic, it does open the door for a variety of parametric inferential statistics to be used. So, there is an advantage. My suggestion is that you regard your items in the way that your field indicates that you should. If you are in a hard science field, ordinal. Behavioral or social sciences, scale. Hope this helps!
- Badges
- Science topic
- Similar topics
- Social Psychology
- Attitude