See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/331950617

01 A note on the usage of Likert Scaling for research data analysis

Preprint · January 2010

CITATIONS READS
0 1,481

1 author:

Jonald Pimentel
University of Southern Mindanao
12 PUBLICATIONS   73 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

collaboration project View project

personal project View project

All content following this page was uploaded by Jonald Pimentel on 22 March 2019.

The user has requested enhancement of the downloaded file.

 USM R & D 18(2): 109-112 (2010)
ISSN 0302-7937
A note on the usage of Likert Scaling for research data analysis

Jonald L Pimentel

Department of Mathematics and Statistics, College of Arts and Sciences, University of Southern Mindanao
Kabacan, Cotabato, Philippines, Email: [email protected]

Abstract

In this paper, biases on some used known method adapting Likert Scaling are shown. Then, an improvement of the method to
correct or minimize the bias is presented in order to have better interpretation of the result of the analysis of the data.

Keywords: bias, bipolar scaling, difference, Likert Scale, psychometric scale, rating scale, statistical tool.

Introduction response levels are used although many psychometricians

suggest using seven or nine levels. The reason for this is
Likert scaling is a psychometric scale commonly used based on a recent empirical study by Dawes (2008) that a
in questionnaires and is the most widely used scale in 5- or 7- point scale may produce slightly higher mean scores
survey researches particularly in social science researches. relative to the highest possible attainable score, compared to
Invented by a psychologist named Rensis Likert, he sought those produced from a 10-point scale, Further, he stressed
to find effective and systematic means of studying human that this difference was statistically significant. On the other
attitudes and the factors that influence them. His research hand, In terms of the other data characteristics, there was
led him to develop a scale for attitude measurement which very little difference among the scale formats in terms of
now bears his name. This term is interchangeably used with variation about the mean, skewness or kurtosis. The format
rating scale even though the two are not synonymous. If a of a typical five-level Likert item is: 1. Strongly Disagree,
respondent wanted to respond to a Likert questionnaire item, 2. Disagree, 3.Neither Agree nor Disagree (undecided or
this respondent specifies a level of agreement to a particular neutral), 4. Agree, 5. Strongly agree.
statement.
Likert scaling is a bipolar scaling method, measuring
Likert scale received a wide range of acceptance. It is used either positive or negative response to a statement. Likert
not only in psychology and marketing researches but also in scales may be subject to distortion from several causes.
other areas: education (Anonymous 1998), medicine (Grant Respondents may avoid using extreme response categories
et al 1999), nutrition (Lindhorst et al 2007), nursing (Seal (central tendency bias); agree with statements as presented
2007), and other fields in finance, engineering, and human (acquiescence bias); or try to portray themselves or their
study. organization in a more favorable light (social desirability
bias). Designing a scale with balanced keying (an equal
The scale is used either as a summated scale or as an number of positive and negative statements) can obviate
individual scale item. When used as a summated scale, each the problem of acquiescence bias, since acquiescence
item is summated to produce an index. Individual scale items on positively keyed items will balance acquiescence on
are used in eliciting latent variables, such as in Structural negatively keyed items, but central tendency and social
Equation Modeling. desirability are somewhat more problematic.

A simple statement in which the respondent is asked to In this study, the author’s interest is presented in the way
evaluate according to any kind of subjective or objective the Likert scale is applied to descriptive interpretation after
criteria is a Likert item. Generally it is a measure of the level it had underwent several stages of summing up and related
of agreement or disagreement on the particular statement. procedures of the central tendency measures.
Odd numbers of response categories (five, seven, and nine)
are usually used, as well as the least popular even number of Scoring and analysis
responses which are called force-number Likert scale. Four to
six numbers of responses are used in this technique. The odd If a questionnaire is completed, each item may be analyzed
number of three response categories can also be used (Jacoby separately or in some cases item responses may be summed
& Matell 1971). However, often in practice, five ordered to create a score for a group of items.

109
An issue on whether Likert items maybe considered an checked to fulfill the strict formal axioms of the model
ordinal level data or an interval level data is a subject (www.rasch-analysis.com/rasch-model-specification.htm).
of disagreement. In most practice, they are considered
ordinal or interval level data, many regard such items as Some applications
ordinal data because at only five levels, one cannot assume
that respondents’ perceived all pairs of adjacent levels as When items are Likert scaled and assumed to have interval
equidistant. On the other hand, often the wording of response measurement, often in practice, the information to all the
levels clearly implies symmetry of response levels about a respondents are summarized in the form of weighted mean. It
middle category; at the very least, such an item would fall then interpreted using an interval which in turn corresponds
between ordinal- and interval-level measurements; to treat to a verbal description. For example, a recent research in
it as merely ordinal would lose information. Further, if the the common errors committed by nursing students in the
item is accompanied by a visual analog scale, where equal clinical exposure (Jamora et al, Table 1):
spacing of response levels is clearly indicated, the argument
for treating it as interval-level data is even stronger. Table 1. Perceived responses on the common clinical errors
among the nursing students, Jamora et al, 2010.
If considered as an ordinal data, Likert responses maybe
displayed in a graph particularly bar charts. Its center Common clinical errors Weighted mean
being median or the mode but not the mean. Its spread
(variability) uses the range across quartiles but not the 1. Wrong dosage 1.00
standard deviation since mean and standard deviation are 2. Wrong time 1.75
inappropriate measures for ordinal data (Jamieson 2004); 3. Failure to do a procedure correctly 2.25
analysis uses non-parametric tests such as the chi-square 4. Failure to check the equipment 2.00
test, Mann–Whitney test, Wilcoxon signed-rank test, or 5. Able to break a sterile field 2.75
Kruskal–Wallis test. If guaranteed by the Central Limit
Theorem that ordinary averages of the Likert scale data
behave normally, parametric analysis maybe performed; Table 1a used a Likert Scale in the following manner: 1.
however, some would disagree that ordinary averages Never, 2. Rare, 3. Sometimes, 4. Often and 5. Always.
should be used instead for Likert scale data. Giving a verbal description, one has to create an interval of
means in order to give interpretations for the weighted mean.
Responses to several Likert questions may be summed Results of most studies treat the problem in this manner.
with the assumption that all questions use the same Likert Some authors construct a legend indicating description that
scale and that the scale is a defendable approximation to runs like the table below.
an interval scale, in which case they may be treated as
interval data measuring a latent variable. Hence, parametric Likert Scale Interval Description
statistical tests such as the analysis of variance maybe
applied. Note these can be only applied when more than five 1 1.00-1.50 Never
Likert questions are summed. 2 1.51-2.50 Rare
3 2.51-3.50 sometimes
Level of measurement 4 3.51-4.50 often
5 4.51-5.00 always
The five response categories are often believed to represent
an interval level of measurement. But this can only be the The table 1 would be then presented as
case if the intervals between the scale points correspond to
empirical observations in a metric sense. We bear in mind Table 1b. Perceived responses on the common clinical errors
that treating ordinal scales as an interval scales has long among the nursing students, Jamora et al, 2010.
been controversial; however, in principle it will be used as a
basis for obtaining interval level estimates on a continuum Common clinical errors Weighted mean Description
by applying item response models. In polytomous Rasch
model, data maybe obtained that fits this model. In addition, 1. Wrong dosage 1.00 Never
the polytomous Rasch model permits testing of the hypothesis 2. Wrong time 1.75 Rare
that the statements reflect increasing levels of an attitude 3. Failure to do a
or trait, as intended. For example, application of the model procedure correctly 2.25 Rare
often indicates that the neutral category does not represent a 4. Failure to check the
level of attitude or trait between the “Disagree” and “Agree” equipment 2.00 Rare
categories. Again, not every set of Likert scaled items maybe 5. Able to break a sterile
used for Rasch measurement. The data has to be thoroughly field 2.75 Sometimes

110
However, the kind of interval given above is biased in the Table 1d. Perceived responses on the common clinical errors
sense that the difference in the upper limit and lower limit of among the nursing students, Jamora et al, 2010.
the first and last intervals is much lower as compared to the
three middle intervals (as shown in the table below). This Likert Scale Interval Difference Description
observation will result to items having more weighted mean
in the middle. 1 1.00-1.66 0.66 Bad
2 1.67-2.33 0.66 Undecided
Likert Scale Interval Difference Description 3 2.34-3.00 0.66 Good

1 1.00-1.50 0.50 Never Likert Scale Interval Difference Description
2 1.51-2.50 0.99 Rare 1 1.00-1.75 0.75 Very bad
3 2.51-3.50 0.99 sometimes 2 1.76-2.51 0.75 Bad
4 3.51-4.50 0.99 often 3 2.52-3.27 0.75 Good
5 4.51-5.00 0.49 always 4 3.28-4.00 0.72 Very good

To correct this problem, the author reduced or eliminated Likert Scale Interval Difference Description
the bias, by making the difference in each interval to have 1 1.00-1.85 0.85 Very bad
a uniform difference; hence, the new interval given below 2 1.86-2.71 0.85 Rather bad
have a constant uniform difference in each interval. That is, 3 2.72-3.57 0.85 Bad
4 3.58-4.43 0.85 Neither good
Likert Scale Interval Difference Description nor bad
5 4.44-5.29 0.85 Good
1 1.00-1.79 0.79 Never 6 5.30-6.15 0.85 Rather good
2 1.80-2.59 0.79 Rare 7 6.16-7.00 0.84 Very good
3 2.60-3.39 0.79 sometimes
4 3.40-4.19 0.79 often
5 4.20-5.00 0.80 always In this manner, descriptive interpretation of the weighted
mean of the item for three, four, and seven response
Thus, Table 1b will have new description for item 2 (from categories will be properly addressed.
rare to never) that is,
References
Table 1c. Perceived responses on the common clinical errors
among the nursing students, Jamora et al, 2010. Anonymous. 1998. The Effect of Right or Left Placement
of the Positive Response on Likert-Type Scales Used by
Common clinical errors Weighted Description Medical Students for Rating Instruction. Education for
mean Health 11: 122.

1. Wrong dosage 1.00 Never Babbie ER. 2005. The Basics of Social Research. Belmont,
2. Wrong time 1.75 Never CA: Thomson Wadsworth. p. 174. ISBN 0534630367.
3. Failure to do a procedure
correctly 2.25 Rare Dawes J. 2008. Do Data Characteristics Change According
4. Failure to check the equipment 2.00 Rare to the number of scale points used? An experiment using
5. Able to break a sterile field 2.75 Sometimes 5-point, 7-point and 10-point scales. International Journal of
Market Research 50 (1): 61–77.

The correction of the bias in the interval making the Grant S, Aitchison T, Henderson E, Christie J, Zare S,
difference uniform changes the descriptive interpretation of McMurray J, & Dargie H. 1999. ‘A Comparison of the
the item. Thus, a recommendation for three, four, and seven Reproducibility and the Sensitivity to Change of Visual
response categories Likert scale “How good do you think Analogue Scales, Borg Scales, and Likert Scales in Normal
was Macapagal-Arroyo as president?” The following may Subjects during Submaximal Exercise’ Chest 116: 1208-
be studied: 1218.

111
Jacoby J & Matell MS. 1971. ‘Three-Point Likert Scales are Meyers LS, Anthony G, & Glenn G. 2005. Applied
Good Enough’ Journal of Marketing Research 8: 495-501. Multivariate Research: Design and Interpretation. Sage
Publications. p. 20. ISBN 1412904129.
Jamieson S. 2004. Likert scales: how to abuse them
Blackwell Publishing Ltd MEDICAL EDUCATION 2004; Mogey N. 1999. So You Want to Use a Likert Scale?
38: 1212–1218. Learning Technology Dissemination Initiative. Heriot-Watt
University. Retrieved April 30, 2009.
Jamora CJD, Jumaway MS, & Ines CC. 2010. Common
Errors Committed by Nursing Students during Clinical Saris WE, Gallhofer I, & Van der Veld W. 2003. A scientific
Exposure. Unpublished Undergraduate thesis, University of Method for Questionnaire Design:SQP. University of
Southern Mindanao, Kabacan, Cotabato. Amsterdam.

Latham GP. 2006. Work Motivation: History, Theory, Seal M. 2007. Patient Advocacy and Advance Care
Research, And Practice. Thousand Oaks, Calif.: Sage Planning in the Acute Hospital Setting. Australian Journal
Publications. p. 15. ISBN 0761920188. of Advanced Nursing 24:29-37.

Lindhorst K, Corby L, Roberts S, & Zeiler S. 2007. Rural Trochim WM. 2006. Likert Scaling. Research Methods
Consumers Attitudes towards Nutrition Labelling. Canadian Knowledge Base, 2nd Edition.. Retrieved April 30, 2009
Journal of Dietetic Practice and Research 68:146-150.
Uebersax JS. 2006. Likert Scales: Dispelling the Confusion.
Likert R. 1932. A Technique for the Measurement of Retrieved August 17, 2009.
Attitudes. Archives of Psychology 140: 1–55.
www.rasch-analysis.com/rasch-model-specification.htm

112

View publication stats