ORDINAL DATA

See also WHY DOES IT MATTER WHAT SORT OF DATA YOU HAVE?

An ordinal measure involves an ORDERED SERIES, such as graded responses to an item on a questionnaire:

EXAMPLE 1

Questionnaire item:

In the past week, I have often had difficulty concentrating.

Response choices:

Strongly agree = 4, Agree = 3, Neither agree nor disagree = 2, Disagree = 1, Strongly disagree = 0.

EXAMPLE 2

Questionnaire item:

Please rank order the following service aspects from 1 to 5 where 1 = most important to me, and 5 = least important:

How close it is to your home  
How the inside of the building looks  
Friendliness of reception  
Privacy  
Qualifications of therapists  

WARNING!

You might think that if you have data that look like the above, you should select ORDINAL, but before you do, answer the following two questions:
  • Does your measure consist of Likert-scaled items? What are Likert scales?
  • Has the measure been standardised? What is standardisation?
    •

    If the answer to one or both of these is YES, you might consider selecting INTERVAL to describe your data even though it is technically ORDINAL. You can use more powerful PARAMETRIC statistical analyses on interval data than you can with ordinal, so this is an important decision. To find out more, read the sections below.

    Likert Scales

    Many psychological scales have items that are rated on a Likert scale. Here, 'the item is presented as a declarative statement, followed by response options that indicate varying degrees of agreement with or endorsement of the statement' (de Vellis, 1991 p. 68). Although it is difficult to say for sure that the intervals on these scales are equal (see INTERVAL data), and therefore scales made up of Likert-scaled items are technically ordinal level, there has been debate about this point, with some people holding that such scales are better than what we would normally think of as ordinal. In relation to using parametric statistics on these scales, 'one view states that, most of the time, providing you have a good quality ordinal measure, you will arrive at the same conclusions you would have using more appropriate tests' (Fife-Schaw, 1995 p. 47) (our italics).

    Standardisation

    According to Coolican (1999, p 157), standardisation is the process of 'adjusting [a] test until scores on it form a normal distribution'. Coolican (P 187) also wrote: 'Attempts are made to convert many psychological scales to interval level using standardisation' (our italics).

    Still having difficulty deciding whether to click ORDINAL or INTERVAL to describe your data?

    Decisions about statistical analysis are not always clear-cut. If they were, statistics might be easier for many psychologists to understand. If you are still wondering what to decide - ordinal or interval - it may be worth consulting someone locally, or reading some of the sources referenced here. Whatever you do, you may need to give a rationale in any report you do, especially if it is to be examined as part of a training programme. It is to be hoped that the material here, combined with further reading or consultation, will help you formulate that rationale, as well as increasing your understanding of the problem.

    If you haven't already, try other links from this page to explore issues further.

    References

    Coolican, H. (1999) Research Methods and Statistics in Psychology, Second Edition. London: Hodder & Stoughton

    De Vellis, R.F. (1991) Scale Development: Thoery and Applications. London: SAGE.

    Fife-Schaw, C. (1995) Levels of Measurement. In (G. M. Breakwell, S. Hammond & C Fife-Schaw (Eds) Research Methods in Psychology. London: SAGE.

    Back to Which Test Home Page