Research Measurements

The 4 types of measurements discussed in previous posts are linked below. An easy acronym that may help you remember the four types is NOIR. This is also convenient in remembering the increasing complexity of measurement types, i.e. ratio is a more complex measurement than interval.

• Nominal
• Ordinal
• Interval
• Ratio

Experimental Design and Measurements

When designing an experiment, it’s important to consider both the analyses which can be performed and will be performed after data collection. As a general rule of thumb, the most complex measurement allowed by the experiment ought to be collected. Of course each experiment is unique and other costs need to be considered in addition to measurement type, but it is important to think about this during the experimental design stage.

Consider an experiment in which people are separated by handedness: right, left or ambidextrous. The nominal (categorical) nature of the measurement does not allow other measurement types. In this experiment the most complex measurement type is nominal.

In contrast, an experiment which unnecessarily categorizes a continuous measurement can alter findings and affect outcomes. As an example, think about an analysis which breaks hours of sleep into 2 categories: more than 6 hours of sleep and less than 6 hours of sleep. This ratio measurement has been reduced to a nominal measurement and has the potential to obscure/exaggerated important findings. The recommended course of action is to collect the hours of sleep as a ratio measurement since it is the most complex measurement type possible.

Collecting the most complex measurement type is beneficial for several reasons:

• Outlier Investigation
  • In the hours of sleep example above, a person who sleeps 18 hours a day vs 7 hours a day are both in the same category. This has the potential to give a false impression about differences when comparing to another category.
• Post-Hoc Analysis
  • It is often the case experiments give rise to further questions not originally considered. Collecting the most complex measurement type up front provides opportunity for further analysis with other methods.
• Meta Analysis
  • Meta Analysis (or an analysis of a multiple analyses) is an important role in modern social scientific research. It’s important to collect as much information as possible to allow researchers to make fair comparisons between experiments at a later date.
• Future Proofing
  • Data analysis methods did not stall out in the mid 1950s, new methods are being created even as you read this article. By collecting the most complex measurement type possible, the opportunity to use these new methods in the future is maintained.

Cheat Sheet

Below is a cheat sheet to aid in learning about these measurement types, downloads are available in pdf or docx. An in-depth examination of each measurement type is also available: nominal, ordinal, interval, and ratio.

Happy New Year From GoFactr

Make sure to use GoFactr.com for all your research statistical needs. The online statistics program that computes descriptive statistics, ANOVAs, T-Tests, Linear and Logistic regression and Factor Analysis. No software to download means it runs conveniently from your tablet, chromebook, smartphone, laptop or desktop.

Comments

Acknowledgements

Photo by falconp4