Is the level of measurement reflected on the method of statistical analysis of the measured variables using Likert scale ?
Hello Hatim,
This is a topic that has been discussed many times, in RG and elsewhere. Just google the phrase, "Is a Likert-type scale ordinal or interval."
In brief:
1. The response scale for an individual item, as typically quantified using integer values, 1 = lowest, 2 = next lowest, etc., is ordinal. No question about it.
2. Use of individual Likert-type as variables is not nearly as good a procedure as is using a set of items (all relating to the same target) instead, via a summated or averaged score. That is the method that Rensis Likert was presenting in his 1932 monograph, A Technique for the Measurement of Attitudes (https://legacy.voteview.com/pdf/Likert_1932.pdf). These scores tend to be both more reliable and behave pretty well as pseudo-continuous scores without further massaging. You will find that many people use them as if they are interval strength. However, if you're the least bit skeptical, you can factor the scores over a set of related items and generate factor scores that will, in all likelihood, represent interval strength scores.
3. What Likert found (and reported in his monograph) was that, the simple, integer scaling of response option values correlated so highly (in the .90s) with values derived from more intensive scaling methods (of the day) that were purported to yield interval scores that he suggested the extra effort associated with the more intensive scaling wasn't worth the time. Of course, in those days, computers and statistical software weren't available.
4. Several IRT models (such as the graded response model) can take Likert-type response scales for a set of questions (pertaining to the same target) and scale them onto an interval-strength scale.
Good luck with your work.
Hello Hatim,
Originally, the Likert items consisted of statements about feelings or opinions (for example, "The school teaches children to help one another"), for which there no wrong or right answer; the response options consists of three, five, or seven classes with strongly agree and strongly disagree (i.e., ordinal) on left-hand and right- hand side, respectively. The "distance" between items is unknown, and might be different for items on the right- hand, and left-hand side.
At the interval level, the scores are expressed in numbers to quantify results (blood pressure, body temperature, etc.). In this case, the distances between scores are known ( distance between systolic blood pressures of 120 and 130; and between 130 and 140 mmHg is the same).
Best wishes,
Tatyana.
Hatim - I think we discussed this question on ResearchGates. Read the responses given they are useful to you. Also, I recommend reading this article.
"Can Likert Scales be Treated as Interval Scales?—A Simulation Study", by
Huiping Wu & Shing-On LeungPages 527-532 | Published online: 06 Jun 2017
- https://doi.org/10.1080/01488376.2017.1329775
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS