I'm getting wrapped around data types and I need some help:

enter image description here

If you look at the picture above (taken from here), it has the data types like this:

  • Quantitative (Discrete, Continuous)
  • Qualitative (Nominal (N), Ordinal (O), Binary(B)).

enter image description here

But if you look at this next picture (from here), the categories are:

  • Quantitative (Discrete (NOB))
  • Qualitative

One picture has NOB under Qualitative, the other has it under Quantitative. Which one is correct?

  • 5
    $\begingroup$ Every single bullet in the description of "discrete data" is wrong and misleading. $\endgroup$
    – whuber
    Jul 4, 2015 at 14:46

5 Answers 5


All, I couldn't find one picture that put everything together, so I made one based on what I have been studying. Putting the scales of measurement on the same diagram with the data types was confusing me, so I tried to show that there is a distinction there.

enter image description here

I appreciate your help and thoughts! Regards, Leaning

  • $\begingroup$ Discrete quantitative variables (like counts) also can be measured using interval or ratio scale! See here, for example. $\endgroup$
    – Rodvi
    Dec 8, 2019 at 12:59

These typologies can easily confuse as much as they explain.

For example, binary data, as introduced in many introductory texts or courses, certainly sound qualitative: yes or no, survived or died, present or absent, male or female, whatever. But score the two possibilities 1 or 0 and everything is then perfectly quantitative. Such scoring is the basis of all sorts of analyses: the proportion female is just the average of several 0s for males and 1s for females. If I encounter 7 females and 3 males, I can just average 1, 1, 1, 1, 1, 1, 1, 0, 0, 0 to get the proportion 0.7. With binary responses, you have a wide open road then to logit and probit regression, and so forth, which focus on variation in the proportion, fraction or probability survived, or something similar, with whatever else controls or influences it. No one need get worried by the coding being arbitrary. The proportion male is just 1 minus the proportion female, and so forth.

Almost the same is true when nominal or ordinal data are being considered, as any analyses of such data hinge on first counting how many fall into each category and then you can be as quantitative as you like. Pie charts and bar charts, as first encountered in early years, show that, so it is puzzling how many accounts miss this in explanations.

Put another way, you can classify raw or original data as first reported and as appearing in say the cell of a spreadsheet or database. But its original form is not immutable. Imagine something stark like a death from puzzlement from reading too many superficial textbooks. That can be written on a certificate, but statistical analysis never stops there. There is an aggregation to counts (how many such deaths in a area and a time period), a reduction to rates (how many relative to the population at risk), and so on.

So, how the data are first encoded rarely inhibits their use in other ways and transformation to other forms. The etymology of data is here revealing: translating the original Latin literally, they are as given to you, but there is no rule against converting them to many other forms.


It depends what you mean by "quantitative data" and "qualitative data".

I think the two sites you cite are using the terms differently. Suppose, for example, you ask people:

Did you vote for Obama, Romney, someone else or no one in the presidential election?

What sort of data is this? The variable is nominal: It's only names, there is no order to it. But many people would call it quantitative because the key thing is how many choose which candidate. That's as opposed to qualitative data which might be transcriptions of interviews about what they like best about Obama (or Romney or whoever).

A better way to look at it is to clearly distinguish quantitative data from quantitative variables.

  • $\begingroup$ In the first case, there is one variable, which holds president-name. The variable is qualitative, to be precise is nominal. In the second case, every president-name corresponds to an individual variable, which holds the voters. If, voter-names are known, and, it holds voter-names, then variable is nominal. If it holds number of votes, the variable is quantitative, to be precise is in ratio scale. $\endgroup$ Mar 8, 2020 at 9:40

Neither of these charts are correct. They are rather nonsensical and you are right to be confused (aside from the contradiction).

They seem to be conflating the ideas of fundamental variable type and variable selection to model a system (with a pdf).

There are 3 fundamental variable types (excluding subtypes): Nominal (categorical/qualitative), Ordinal, and Continuous (Numeric, Quantitative). Ordinal has both a qualitative and quantitative nature.

Attribute is not really basic type but is usually discussed in that way when choosing an appropriate control chart, where one is choosing the best pdf with which to model the system. This is sometimes called "attribute data", but it's type is nominal (aka categorical etc). Like Nick mentioned, we count nominals, so it can be confused with a numeric type, but its not.

  • 1
    $\begingroup$ Mandata, based on what you are saying, what changes would you make to the chart I made above? i appreciate your help. Regards, Leaning $\endgroup$
    – Leaning
    Jan 28, 2016 at 16:19
  • $\begingroup$ Mandata, all these charts from different experts are partly correct. The thing is that people understand words and concepts not fully identically but they prefer, for some long or short time, to stack to their own comfortable understanding. For example, some people will reject to call ordinal scale "quantitative" while other will accept, depending of whether "quantity" is necessarily manifest of potentially underlying category of being. $\endgroup$
    – ttnphns
    Jan 28, 2016 at 17:37
  • $\begingroup$ @Leaning. That chart is better than your last one. I would consider discrete a quality of type, not a type itself. Nominal and ordered are entirely discrete, while countable (finite or infinite) quantitative is also. Binary is also a characteristic of type (it is a subset of discrete). Binary is rarely ordered, and almost always is represented by nominal variables. Categorical and nominal are synonyms. difference between ordered variables are hardly meaningless, they may be partially or entirely unknown, or not relevant (the latter implies meaninglessness), but I would not assert that. $\endgroup$
    – mandata
    Feb 2, 2016 at 3:39
  • $\begingroup$ @ttnphns, I agree with what you are saying in spirit, but they both have serious conceptual errors. the first mixes the idea of attribute data type, which is used in selecting a control chart, which basic data type. The second has nominal as a subset of discrete which is a subset of continuous. I might subset discrete, but nominal belongs under qualitative. Maybe its there because one counts nominal events discretely, but even if that is why it is incorrect. $\endgroup$
    – mandata
    Feb 2, 2016 at 3:43
  • $\begingroup$ I don't feel the Interval / Ratio theory is a valid way of describing variable type. It might be good for determining what functions are reasonable when one does not feel confident about the math, but beyond that, I see one scale as a transformation of another scale if they represent the same dimensions or units. $\endgroup$
    – mandata
    Feb 2, 2016 at 3:49

I found this question while searching about levels of measurement and related concepts. I think the charts in the question lack the context. When we do the categorization we define the rules for grouping the objects according to our purpose. So what is the purpose? And are we talking about the variables?

We could categorize variables according to the levels of measurement, then we could have 4 scales (groups) with the following rules:

nominal: attributes of a variable are differentiated only by name (category) and there is no order (rank, position).
ordinal: attributes of a variable are differentiated by order (rank, position), but we do not know the relative degree of difference between them.
interval: attributes of a variable are differentiated by the degree of difference between them, but there is no absolute zero, and the ratio between the attributes is unknown.
ratio: attributes of a variable are differentiated by the degree of difference between them, there is absolute zero, and we could find the ratio between the attributes.

And this is only one approach from Stanley Smith Stevens. There are several other typologies.

Continuous and discrete variables are mathematical concepts where we have a range of real numbers and:

continuous variable can take any value in this range. The number of permitted values is uncountable.
while for discrete variable the number of permitted values in the range is either finite or countably infinite.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.