Eye colour and portfolio types

In any introductory statistics course, you'll learn that the four types of data are nominal, ordinal, interval, and ratio. You'll be assessed on this with a multiple-choice question: What data type is Celsius? Then promptly forget this information, probably for the rest of your life.

Unlike most aspects of statistics, the names are instructional. Nominal means no order. Ordinal means there is order, but the gaps have no meaning. Interval means there is order and the gaps are equal. Ratio means there is order, the gaps have meaning and zero means what you'd expect (nothing).

Whilst this is in theory useful, few, if any people will ever use analysis techniques for data other than ratio. I wish the notes I taught to had the following diagram.

Think of types of data like parts of speech - you can live your life without knowing them and muddle along just fine. You might look stupid if you talk to someone who knows these and use them incorrectly. If you're keen to read up in more detail Statology, Statistics by Jim and GraphPad have more examples on this topic.

Another way to think about data types is that if the data is in an SI unit (of which money is the 8th SI unit), you're using ratio, and you don't need to consider the limitation.

If I asked you to analyse eye colour and portfolio types the same way, you'd recoil in concern and question why I have "Dr" in front of my name. They're both nominal data types. Any projection you put on portfolio types, you should apply to eye colour.

It's worth considering that any order forced upon a nominal data type is subjective and likely to influence how you think about the data. Just because you care about the data, doesn't infer there's meaning to the order if it's nominal.

Once you see this, you'll never un-see this.

Why is this hard to implement?

You're unlikely to be undertaking the analysis, therefore you're reviewing the results. The default expectation when reviewing results in a corporate context is to 'lead with the outcome'. The start of any presentation is explaining how important the information they've found is, which is often complemented by an overly complicated visual representation of the data.

No one wants you to ask questions about data types, least of all the people presenting the analysis who probably do not know the answer to your question or the implications. You're in an awkward predicament no matter what you do. You either derail the entire meeting or sit there wondering why people are making decisions about information that's incorrect.

Now you know the problem - what do you do?

The standard education approach would limit the number of staff with an 'Excel licence' to use the software (think pen licence). That's a draconian way of approaching the topic. There's an argument for running a training program on introductory statistics for your staff - this can be fun speaking from experience, and by all means I'll run the training for you.

Another approach is to find someone in the organisation who knows data types and the implications of how these influence data analysis, and have them as a mentor to anyone presenting. Unfortunately, that might end up being you in the first instance, or it could be me.

Once the analysis that is being presented is of such high quality, you'll need a lot less of this work being done and decisions will be easier to make. There are few investments that bring a greater return to your business than using data correctly.