0
$\begingroup$

I read about the multivariate and high-dimensional data set. I found that the multivariate data is the data with more than 3 variables. In addition, the high-dimensional data is the data with a large number of variables. So, for me, the multivariate and high-dimensional data set are the same, is that correct? I think, do not know if it is correct or not, the multivariate data is a data with more than 3 variables but the number of observation is quite small (300 to 500 observations). However, high-dimensional data is the data with a large number of variables (10 to 100 or even larger) and a large number of observations (1000 observations). Is that correct?

$\endgroup$
2
  • $\begingroup$ multivariate data is the data with more than 3 variables No. Please read the tag [multivariate-analysis] info. As well as [high-dimensional] tag. $\endgroup$
    – ttnphns
    Jul 1 '19 at 9:18
  • $\begingroup$ @ttnphns thank you for your comment and I am really sorry if I use a wrong tag. I actually have to work with multivariate and high-dimensional analysis. $\endgroup$
    – Mary
    Jul 1 '19 at 10:08
1
$\begingroup$

What is considered 'high' in high-dimensional is a futile discussion in my opinion. However, univariate, bivariate and multivariate usually refer to the number of outcome variables.

High-dimensional data analysis is a term more often used for dealing with things specific to a large number of explanatory variables, or parameters. For example:

  • Inability to perform matrix inversion and most forms of decomposition when $n < p$;
  • Increasingly sparser coverage of the sampling space with growing dimensions;
  • The combinatorial explosion

A multivariate approach, can but need not involve these issues. You can have multiple outcome variables with only a handful of explanatory variables. You can also have a high-dimensional multivariate problem.


Tangentially related, for a classification task consisting of two classes, the outcome can be argued to follow a binomial distribution, whereas a multi-class classification task can be argued to follow a multinomial distribution.

$\endgroup$

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.