### Data and Levels of Measurement

One of the trickiest parts of the G544 approaches and research methods in Psychology exam is making sure you follow the directions provided with the question. Many people have problems with directions regarding levels of measurement, for example: ‘use ordinal data’. Hopefully this article will help you understand the different levels of measurement differently and how they can be collected, which will be crucial in ensuring you can easily answer all the questions in the exam.

There are four different levels of measurement:

**Nominal****Ordinal****Interval****Ratio**

##### Nominal Data

This is by far the simplest type of data. Quite simply, **Nominal Data** is data which **cannot be assigned a numerical value of any true mathematical significance. **

Let’s look at an example.

Suppose we have four participants in an exam:

Participant 1

Participant 2

Participant 3

Participant 4

This is nominal data because the name assigned to each participant, while numerical, does not have any mathematical significance. From these names alone we cannot rank the participants because they simply have been assigned a random number. Participant 1 could easily swap names with Participant 3 without changing anything.

Let’s look at another example.

Suppose again we have four exam participants, but this time we will not assign them a random number, but instead collect their names.

Jeff

Alan

Amy

Ludwig

Again this time the data is Nominal because the data is simply a name and nothing more.

##### Collecting Nominal Data

Firstly, as nominal data is simply a name we cannot use any measures of central tendency, such as the mean, mode and median with this type of data. We cannot work out what the mean name of exam participants is, nor would there be any use in doing so.

Let’s look at a potential study that could collect nominal data.

Suppose you want to study happiness and sadness on a Monday morning. You decide to use a questionnaire to collect this data. This could be a very simple questionnaire with one question: ‘Please circle the word which best describes your mood on this Monday morning.’ Answers: ‘Happy’, ‘Sad’ – This would collect nominal data.

The results of this study might be:

Happy – 12

Sad – 38

We could display these results in a number of ways including bar charts and pie charts.

Depending on what method and design you have used you will have to use different inferential statistics to analyse the data.

- For experiments using Independent measures: Chi-squared
- For experiments Repeated measures: Sign Test
- For correlations collecting nominal data: Chi-squared

##### Ordinal Data

**Ordinal data** is data which is numerical and can be **put into an order,** but nothing else can be inferred from the numbers, we can only use the numbers to establish an order.

We can use measures of central tendency and measures of dispersion with ordinal data.

Let’s look at an example of ordinal data.

Suppose they are 5 people competing in a running race. We can rank their positions when they finish the race. This is ordinal data.

1st, 2nd, 3rd, 4th, 5th

These ordinal data only tells us the order in which the participants finished in, it does not tell us anything else such as the time it took each runner to complete the race. Furthermore, we cannot establish the differences between the runners in the race, that is to say, we cannot know if 1st and 2nd were separated by a few seconds or by a few minutes.

Let’s look at another example.

Suppose you are trying to understand people web browsing interests. So you use a questionnaire which asks the participants to rank their top 5 favourite websites out of a list of 5.

- Google.com
- Psychyogi.org
- Wikipedia.org
- Reddit.com
- Youtube.com

As you can see, we can rank this data, but again, because it is ordinal, we cannot know the differences between the websites, we don’t know if google was our participants clear favourite or whether it Psychyogi.org came in a close second.

##### Collecting Ordinal Data

If the exam question stipulates that you should collect ordinal data, then the easiest way of doing that is to use a self-report questionnaire with rating scales. Remember that you can use questionnaires within experiments, if the question also stipulates that you have to use an experiment. You can also use structured observations because they collect ordinal data.

Let’s look at an example.

Suppose you want to study the perceived motivation to pursue a fitness regime. So you decide to use a self-report questionnaire in order to research this. Some of the questions you could include could be.

‘On a scale of 1-10 please rate your motivation to complete this fitness regime (1 being not at all and 10 being completely motivated).’ Answer: ‘1 – 10’

‘On a scale of 1-10 please rate your self-efficacy (belief in your ability to achieve) in regards to this fitness regime (1 being not at all and 10 being I can easily complete this fitness regime). ‘Answer: ‘1-10’.

‘Do you believe that this fitness regime will lead to a better quality of life’ Answer: ‘Yes – No’

The first two questions use scales and collect ordinal data and the third is simply a ‘yes no’ question and it also collects ordinal data.

Let’s look at an example of using a structured observation to collect ordinal data.

Suppose you want to study the behaviour of a group watching a television show. You generate a coding system that establishes what you are going count as behaviour. When you complete the observation you give each behaviour a **rating. **

Giving the behaviours a rating is the most important part of collecting ordinal data from structured observations.

Our results from this study may look something like this:

Scores out of ten rated the intensity of behaviours during a viewing of a hour long television show.

Laughing – 9

Playing with phone – 7

Talking – 4

As you can see we data we have collected is ordinal from this structured observation.

Depending on what method and design you have used you will have to use different inferential statistics to analyse the data.

- For experiments using Independent measures: Mann-Whitney U Test
- For experiments Repeated measures: Wilcoxon Test
- For correlations collecting ordinal data: Spearman’s Rank

##### Interval and Ratio Data

Interval and ratio data is ordinal data, but better. The reason that both interval and ratio data are better than ordinal data is because we can tell the differences between the scores with both interval and ratio data, we cannot do that with ordinal data.

Interval data has data points which are evenly spaced. The simplest example of interval data is temperature because the difference between data points is always the same. For example 1˚C and 2˚C have the same distance between them as 60˚C and 61˚C.

Ratio data is more precise than interval data because it has an absolute zero. One of the most common examples of ratio data is time, because 0 seconds really is the lowest amount of time, unlike temperature, which does go into negative numbers.

##### Collecting Interval/Ratio Data

Let’s suppose we want to collect interval or ratio data. Then one of the simplest ways we can do this is through the administration of a standardised test such as an IQ test or we could administer a timed task.

For example the results of five participants on an IQ test could be:

- 131 IQ
- 117 IQ
- 101 IQ
- 101 IQ
- 92 IQ

Depending on what method and design you have used you will have to use different inferential statistics to analyse the data.

- For experiments using Independent measures: Unrelated-t test
- For experiments Repeated measures: Related-test
- For correlations collecting interval ratio data: Pearson Product Moment

##### Further Reading