I see a lot of references to hierarchical data but what exactly counts as hierarchical data isn't always clear and I can't find a clear definition of what counts as hierarchical data.

For example, the following blog states:

"[neural network are]...particularly suited to complex, hierarchical data"


A few examples are give involving images, etc, but can hierarchical data also include data who's independent variable is continuous?

How would you define hierarchical data, would you be able to provide an example?

  • $\begingroup$ I'm voting to close this question as off-topic because this is a definition which is not a topic for discussion here. $\endgroup$ – Michael R. Chernick Feb 7 '17 at 14:16
  • 4
    $\begingroup$ @MichaelChernick I am not aware of any such criterion, that definitions are off-topic. What is on and off-topic is defined in stats.stackexchange.com/help/on-topic For example, we have a tag "terminology" and many threads under that tag. There is a clear statistical and/or machine learning interest in different kinds of data; other people outside those fields might have definitions but in my view it is sufficient for there to be interest here to keep the question open. $\endgroup$ – Nick Cox Feb 7 '17 at 14:42
  • 3
    $\begingroup$ @MichaelChernick, definitions are potentially on topic. We even have a definition tag w/ 185 threads. $\endgroup$ – gung - Reinstate Monica Feb 7 '17 at 16:00
  • $\begingroup$ @kjetilbhalvorsen I am removing [hierarchical-data] tag because I don't see how it's different from [hierarchical-analysis] which we have as a synonym of [multilevel-analysis]. No need to create a new tag for this question. I am writing this because I noticed that you added this tag back after I removed it yesterday (so I removed it again now). $\endgroup$ – amoeba says Reinstate Monica May 15 '17 at 11:14
  • $\begingroup$ I do think I added it by an error. So good you removed it! $\endgroup$ – kjetil b halvorsen May 15 '17 at 11:23

Data being hierarchical does not depend on what type the variables are, but on the structure of the data. If individual observations are grouped into groups or clusters, then your data is hierarchical.

For instance, if you are measuring children's test results, you have to account not only for variations between children, but for the effect of the teacher on the children, the school, the school district, the city, etc. So, in this example, the hierarchy is child - class - school - district - city. Or if you are studying the effect of a medical treatment, patients are the individuals, but they belong to different hospitals, then the heirarchy is patient - hospital.

I hope it helps.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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