Analysis of Likert Scale Questions

Home Forums Default Forum Analysis of Likert Scale Questions

Viewing 13 posts - 1 through 13 (of 13 total)
  • Author
  • #5764

    Can we use likert scale for correlation and chi-square test? Is is ordinal scale or interval scale?

    Annie Pettit

    Well, in the real world (business), you certainly can and it’s a common occurrence. Likert scales are usually ordinal scales.

    Nick Stone

    Hi George

    I think this interesting question has divided responses. It would depend on who you are trying to convince of a particular case. One argument is that while the scales are strictly ordinal, there is evidence that people (at least in business research) do respond in patterns that are close enough to approximate interval level (personal discussion with Joe Hair, also see his reader friendly texts).

    The other school of thought thinks this is unacceptable and you should either treat them as ordinal, e.g. use non-parametric approaches or transform the data into a more suitable form using something like Rasch analysis or item response modelling.

    Hope this helps



    Thank your very much. Now it is clear
    George Mathew


    I believe there are two camps. One believes that Likert scale data is Ordinal scale, other believe that this data belongs to interval scale. Both camps have published their research findings and analyzed data accordingly. Its up to you to which camps you will follow.


    Hi Georges
    I’m agree with the collegues that a likert scale should be considered as an interval or ordinal scale, but I think, we can also consider the different levels of a likert scale (normally 5 or 7) as a modality of a nominal variable, and then, we can use chi-square test, Multiple Coresspondance analysis …


    You can simply use excel to find the averages and then compare these for each class of questionaire


    im happy to join this website. i will forward to you some some questions regarding the construction and analysis of a five point likert scale. thanks!

    Jeremy Miles

    Most people treat them as interval, although strictly they are ordinal.


    It will be highly appreciated if the claim, ordinal or interval, is associated with sufficient reasoning at this level of discussion.


    As far s i’m depend on your research Question and Research Objectives.From there then you can determines what the right measurement and method for tthe research.


    In Research that seeks to determine the impact of government’s policies on poverty reduction over a 20 year period, what are the two major variables-dependent and independent?
    Thank you
    Peter Betiang


    Spearman’s rho, a rank order correlation coefficient, may offer you a good option for correlation analysis, since it is a nonparametric measure.

    When I worked in a large advertising firm (prior to entering the academic life), we were careful to treat Likert scaled responses as ordinal because the propensity to move from neutral to agree or disagree tending to be greater than movement to the polar choices. (And of course this varies across individuals.) Because it was helpful to be able to take averages, we would score responses to reflect the likely spans. For example, you could score strongly disagree to strongly agree as -5, -2, 0, +2, +5, with an average near 0 indicating little interest in the product.

Viewing 13 posts - 1 through 13 (of 13 total)
  • The forum ‘Default Forum’ is closed to new topics and replies.