LIKERT FREQUENCY SCALE SERIES OF QUESTIONS

It refers to questions that have answer choices that are on an ordinal scale. These days, Likert scale is used more generally. Likert scale originally meant a quite specific type of scale and, until about 20 or 30 years ago, you would frequently see Likert type scale or some variation on that.

The Likert scale originated with Rensis Likert ( 21), and has a long history of use in Kinesiology research ( 13, 14, 24).A Likert scale example for frequency would be something like: Never. AGREEMENT Strongly Agree Agree Undecided.Likert, and Likert-type, responses are popular psychometric item scoring schemes for attempting to quantify peopleвЂ™s opinions on different issues. The bold face sets are the most popular.

If the data are interval, then parametric statistics can be used. If the data are ordinal, then non-parametric statistics are typically considered the most appropriate option for analysis. 7-point Likert scaleThe five point scale that is used here is the most common, but some Likert scales have 4 points , where the not sure category is taken out (indecisive.The long-running issue with Likert-type scales and ordinal responses is the appropriate statistical treatment of these data. Two common alternatives are the 7-point Likert scale. However, the Likert scale doesn't have to use just 5 options.

Chances when I could have. 3 Occasionally, in about 30 of the. 2 Rarely, in less than 10 of the chances when I could have. Whereas the classic Likert-scale items had 5 possible responses, the RPE scale as 14 choices ( 3) and the modified RPE has 10 ( 4).Frequency 7 point. For example, investigators have published research on the Rating of Perceived Exertion ( 3, 4), with almost all treating these data as interval rather than ordinal ( 1, 2, 10вЂ“ 12, 15).

Note that the above response options are вЂњbalancedвЂќ in that the items to the left of вЂњNeutralвЂќ have an equal number of counterparts to the right of вЂњNeutralвЂќ. The purpose of this Research Note is to review current thinking on the treatment of data generated from Likert-type, and other ordinal responses and provide evidence for using alternatives.In a Likert-response item with choices varying from вЂњStrongly DisagreeвЂќ to вЂњDisagree to вЂњNeutralвЂќ, to вЂњAgreeвЂќ to вЂњStrongly AgreeвЂќ, it would appear to be in the mind of the research participant whether or not there is an equal distance between each of these choices ( 9). Most importantly, the goal of research is to produce valid results useful for advancing the field, and valid statistical conclusions require valid statistical analyses.

In fact, Knapp points out, any ordered non-linear numerical assignment, 3, 11, 17, 23, 31 could also be made and preserves the ordinal nature of the data however, this latter non-linear choice would have an impact on group means and whether or not parametric statistics should be used ( 19).But, as Knapp illustrates, if the terms вЂњnever, seldom, occasionally, alwaysвЂќ were used, the two middle values could be argued as being very similar, with perhaps much less distance between вЂњseldomвЂќ and вЂњoccasionallyвЂќ than between вЂњneverвЂќ and вЂњseldomвЂќ, or between вЂњoccasionallyвЂќ and вЂњalwaysвЂќ. Knapp further argues that other numbers could be assigned such as 1, 3, 5, 7, 9, or any other linear transformation, and this would not impact the data or its analysis. If a response has choices, вЂњStrongly DisagreeвЂќ, вЂњDisagreeвЂќ, вЂњNeutralвЂќ, вЂњAgreeвЂќ, and вЂњStrongly AgreeвЂќ, Knapp suggests that these could readily be assigned numerical values of 1, 2, 3, 4, 5, as is often done. For example we might expect respondents to be very sensitive to the change between вЂњRestвЂќ (RPE = 6) and вЂњFairly LightвЂќ exercise (RPE = 11) to be a larger difference than the difference between вЂњHardвЂќ (RPE = 15) and вЂњMaximumвЂќ (RPE = 20) ( 4).Knapp gives a useful illustration of the potential problems of Likert responses which could also be applied to RPE responses.

Likert Frequency Scale Series Of Questions

They do make the important distinction between вЂњLikert ScalesвЂќ compared to the answers to individual questions using Likert-type responses. They suggest the following possibility for the means of a coded 5-item Likert-type response to a series of Questions:Support of Likert Responses as Interval dataCarifio and Perla ( 5, 6) are among the strongest supporters for treating Likert-type responses as interval data, going so far as to suggest that the Likert-responses approximate ratio data. But, after these descriptors are converted to numbers, investigators are comfortable doing just that in fact the results might be (improperly) expressed as вЂњStrongly Agree.523вЂќ.Clason and Dormoody ( 7) offer another critique of Likert response analyses. ( 20) made the observation that no investigator would express the mean of a Likert-response item as вЂњStrongly Agree and a halfвЂќ. With RPE, there is less ambiguity, but it is likely that the lower parts of the scale are further apart than the upper parts of the scale, especially for those less experienced with very hard exertion.Kuzon et al.

This unified result (scale) will be more reliable and reflect the underlying construct better than will any individual item. In their view the individual items in a вЂњscaleвЂќ are not independent and autonomous, but rather must be connected in such a way as to yield a single unified result. Any particular item comprising this scale can have a response format which might or might not be a Likert-type response.Carifio and Perla ( 5, 6) also argue that much of the criticism of вЂњLikert ScalesвЂќ confuses the response format from the actual multi-component measurement (i.e.

Are typically one-question measures and analyzed individually, so the six-or-more-item requirement is violated ( 5, 6).Vickers ( 25) noted greater reliability of the Likert response compared to VAS. Of course research measures of exertion or comfort, etc. Surprisingly, Carifio and Perla ( 5, 6) also tout Vickers ( 25) as having made a strong case for the advantages of the Likert-type response assessment even though the Vickers study only used a one-item survey of pain, and not a proper вЂњscaleвЂќ by their definition given above ( 5, 6). Consequently, Carifio and Perla ( 5, 6) make a strong argument against the statistical or interpretive analysis of individual responses, suggesting that the summative assessment of a series of items is the proper item of analysis and that such a summative assessment yields interval or ratio data.

Regardless of where one stands on the use of F-tests of LikertвЂ“scales or other non-equal interval measures, in any situation in which Pearson correlation-based analyses are planned, then using a VAR, or other alternative, seems to be a more conservative approach with no clear reason for not using such a scale.In the end, it seems the most important thing to keep in mind, is that statistical analyses are not an end in themselves, but rather a means to an end. This conclusion seems baffling when there was no criterion measurement ( 23).Despite their strong support for Likert-scales (as opposed to individual Likert-type item, or, in the kinesiology case, other unequal-interval response), Carifio and Perla concede that Pearson correlations and statistical derivatives (multiple regression, factor analysis, multivariate ANOVA, and discriminant analysis) are not very tolerant of uses of ordinal data, whereas F-tests generally are robust with regard to ordinal data ( 5, 6, 19). In a similar fashion, Vickers ( 25) reported that the Likert-type response to their single question of pain yielded a higher mean value than the same question posed to the same group using a VAS, and concluded that this meant that the Likert-type response was вЂњa more responsive measureвЂќ.

Whether or not a balanced scale is viewed as an interval scale, alternatives to the Likert scaled, and similar items are available. Conversely in situations with very large sample sizes, statistical power can be so high that impractically small changes (effects) are statistically significant but not of meaningful (practical) importance.It seems indefensible to offer an unbalanced Likert scaled item, or any other single-measurement item as an interval measure, especially when other measurement options are available. For example, when studying elite athletes, sample sizes may be small, but small effects may have great practical significance for this population, but the probability of making a Type II error is large. This approach does raise our awareness of Type I and Type II statistical errors. Statistics are not a substitute for thinking about what data truly mean, and what data are showing about the population.Along these lines, Hopkins ( 17, 18) is known for insisting that effect sizes be presented along with p-values.

The VAS typically has descriptive anchors only at the two extremes, although there has not been any published research on VAS with multiple anchors.