Laurie,'Making Users Choices Easier' is simply brilliant.

One of the principal reazons is the level of "sensibility" of your scale to capture the details of the attitudes or reaccions of your participants. In this case, more posibilities may add "sensibility". However, always depends on study object and the questions.

Some questions, based on its nature, could describe properly with less alternatives for answer.

--

Una de las razones fundamentales puede ser el nivel de "sensibilidad" de la escala a las actitudes o reacciones de los participantes en la encuesta. Más posibilidades pueden agregar más sensibilidad. Sin embargo, depende del objeto de estudio y los reactivos diseñados siempre.

Algunos reactivos, por su naturaleza, pueden quedar bien descritos con pocas posibilidades de respuesta.

Yes, it is a good choise and is more commum to use the 5 or 7 levels. I think it is easier to use an impar scale has the respondent can mark easily the meadle of the scale and have a better position to anweer.

Dr. Jose, I strongly agree with you that many researchers choose number of scale categories in Likert-type scale without reporting any rationale for their choice. However, the decision should not be completely arbitrary.

Since the introduction of Likert-type scale by Likert (1932), there has been a considerable number of research works done on finding the number of scale points that maximize reliability. No consensus has so far been reached regarding the optimal number of scale points to be used for maximum reliability. If you really want to be rational in choosing the response categories, you should consult these papers:

• Preston, C. C., & Colman, A. M. (2000). Optimal number of response categories in rating scales: reliability, validity, discriminating power, and respondent preferences. Acta psychologica, 104(1), 1-15.
• Boote, A. S. (1981). Reliability testing of psychographic scales: Five-point or seven-point? Anchored or labeled ? Journal of Advertising Research, 21, 53-60.
• Matell, M. S., &  Jacoby, J. (1971). Is there an optimal number of alternatives for Likert scale items? Study I: Reliability and validity. Educational and Psychological Measurement, 31, 657-674.
• Schutz, H. G., & Rucker, M. H. (1975). A comparison of variable configurations across scale lengths: an empirical study. Educational and Psychological Measurement, 35, 319-324.
• Cicchetti, D. V., Showalter, D., &  Tyrer, P. J. (1985). The effect of number of rating scale categories on levels of interrater reliability: A monte carlo investigation. Applied Psychological Measurement, 9, 31-36.

Hi,

the main guides for researcher selection of the tool are the purpose of the study,nd a the valid and reliable tool. the number of steps in the tool or the number of the items is not the main issue. for likert scale, research usually calculate the total scores from all the items of the scale to change it from ordinal level to ratio level.

Regards

Here is an article that discusses a number of issues related to Likert scales, including what difference it makes to use more or fewer points in the scoring.

Dear David and Khan,

thank you very much for the references. José

I totally agree that the purpose of the study is the main issue. However, in turn, to capture the details of the attitudes/perceptions of the respondents in my opinion is not so easy. It depends on other facts as well, such as the questionnarie design, use of the paper form instead of on-line survey, etc. My phd. was about "Collaboration in Supply Chain". My concern was in understanding the term "Collaboration" and its impact of "logistical performance indicators". I used many variables to summarise those constructs. In order to capture those meaning, I tested 3 scales (1-5; 1-7 and 0-10) with 15 participants and the results showed us that the 0-10 scale was more suitable than the other ones. The respondents felt more confortable (more natural) and faster to understand this range, maybe because in Brazilian culture this scale is very known, instead of 1-7 scale for example.

Thanks everyone for your collaboration.

Jose this is why you can researcher bias,usefully!

I have always preferred using an even scale to ensure that respondents take a view and express and opinion. An odd scale of 5 may result in many respondents putting a 3 thus invaliding any useful  conclusions. Trust this helps

0-10 has been cross-validated as most similar across countries. Culturally, the notion of a none/zero/never and an all/always/completely concept are much more universal than sevens or fives (especially when they lack a zero value).

To avoid midpoint madness, word a question of agreement, intensity, frequency, duration, something for which 0-10 or a categorical scale (never, occasionally, half the time, usually, always being an example) in which the respondent cannot use the midpoint without having the feeling of having made a declarative statement, rather than being able to use it as a refuge from a real answer. The latter most often occurs when agreement or satisfaction statements (sic) are overly abstract and/or general, and when statements of fact are given full-boat scales when they should be small and categorical.

Answers like "neutral" or "neither agree nor disagree" are costly covers for poor design. "Not sure" is a legitimate  response in some cases, as is "did not use" or "can't evaluate yet," but "don't know" is not, unless the question is knowledge. For simple things like did you get a receipt, did the clerk leave a favorable impression, we torture respondents with 5, 7 or 11-point scales when all they need is binary + don't remember. The further we move away from real life and into research-speak, the  further we are from truth.

Dear Laurie and Arup, thank you very much for sharing these concerns. Jose

Laurie,'Making Users Choices Easier' is simply brilliant.

Using 0-10 scale in the questionnaire, normally we have two options:

1. To ask direct to respondent which value (from 0 to 10) they think that is more suitable or value some specific question. In this case the respondent just give us the number without take look at the rank (0-10). It is easy because they now what this scale means.

2. To show the rank, from the lowest value (0-on one side) to the highest one (10-on the opposite site).

I've noticed that, the second option is more efficient, although the respondents take more time to give the response when compared with option 1.

Do you agree? Do you have any other suggestion to share? Thank you so much. José.

Dear José e Rosane,

We can always transform a given scale in another. It  depends on objectives of the study. However, in any one of them should make a descriptive analysis for the exploration of data (mode, outliers, etc).

The choice should also consider the statistical model on which it will apply the scale.

Have a nice time

Helena

Dear Helena, thank you very much for this answer. I do agree it is very important to know the data first through descriptive analysis, and then to apply others statistical methods.

One concern, some authors do not agree to make some descriptive analysis (for example Mean and Std Deviation) in ordinal data using likert scale. I understand this when we have low rank, for example 1-5. However, when we have more possibilities, for example 0-10 or 1-7 scales, I understand that the use of the Mean and Std Deviation are more suitable, isn't it?

Kind regards, José

Dear Jose, 

In a certain way i agree with you, but the Mean of a distribution is only suitable if it is near the Mode.

It is better categorize  the continuous variables at the end of an application in order to a bettter explanation of the results. Then i complement the analysis with models  for categorical variables.

Kind regards

Helena

Dear Helena, Thank very much for the valuable comments! Very nice remark and suggestions, very useful as well. K. regards, José

Dear Jose,

There are some studies which suggest that 7 point likert scale is better compared to 5 point scales in case of non facilitated surveys (like online surveys).

Refer the following article:

Finstad, K. (2010). “Response Interpolation and Scale Sensitivity:

Evidence against 5-Point Scales”, Journal of Usability Studies, 5(3),

104-110

Regards

Asish

Dear Asish, thank you very much for this. I will read. Jose