As I assume, we more often have datasets with (A) very many cases (objects) and not so many variables (properties) (= large data) than the other way around: (B) just a few cases and many more variables (= high dimensionality).

There are other conceivable cases: (C) few cases, few variables, (D) very many cases, equally many variables.

I assume, that for these different cases, different statistical methods will be applied. Is this true, and how can these be characterized?

Further, I wonder, how often each of these cases does occur in practice (proportionally, estimated) - and from which disciplines they typically stem.


closed as too broad by rolando2, Michael Chernick, gung, mdewey, whuber Jan 19 '17 at 19:22

Please edit the question to limit it to a specific problem with enough detail to identify an adequate answer. Avoid asking multiple distinct questions at once. See the How to Ask page for help clarifying this question. If this question can be reworded to fit the rules in the help center, please edit the question.

  • 2
    $\begingroup$ Even though I think the gist of this question is interesting, it is really hard to truly answer it. I can very easily 'guestimate' all four scenarios happen equally often, which may be correct in my field, and wrong in another or in the grand total of studies and data-collections. Moreover, statistical methods aren't applied because of data-structure, but because of relevancy and applicability to the research question, i.e. reason for collecting the data in the first place. Could you clarify what it is you are after? (general discussion, polling which scenario is applicable, ..., etc.) $\endgroup$ – IWS Jan 19 '17 at 14:06
  • $\begingroup$ Would you mind to tell me what your field is? And what means "few" and "very many" in your field? $\endgroup$ – Hans-Peter Stricker Jan 19 '17 at 14:20
  • 1
    $\begingroup$ for what it's worth: biomedical research. As to 'few' and 'very many', I do not mean to be rude but 'few'='few' and 'very many'='very many'... I would not dare generalize towards these vague descriptions for all biomedical studies. Again it is the study goal which directs whether your sample size and 'dimensionality' are relatively small or large. $\endgroup$ – IWS Jan 19 '17 at 14:46

This question may be too broad to answer. What model to use depends on the hardware software and the requirement from business. In addition, as mentioned in the comments, different fields may have different definitions on how big the data is.

  • In medical study, it is very costly to get few data points. If we have thousands patient's data with different treatment, that would be considered is big enough.

  • In bioinformatics, we usually have many "features" (say from DNA) but few "instances".

  • In computer science, say pattern recognition on images. It is very common to have billion of instances (pictures) and millions of features (pixels in an image) for one instance.


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