1

I have a data set including 25 variables $(x_{1,t},\dotsc,x_{25,t})$ at each time $t$ and all of this group is repeated through time. I want to explore the relationship between these variables through time and for example say that $x_{1,t}$ is highly correlated with $x_{t,10}$ during lapsed time. Then summarize the number of variables so that those which are more crucial and important be selected.

I am not looking for forecasting but I just to find out which variables are highly correlated and then decrease the number of variables.

I have read about factor analysis but I think it is not applicable for time series data.

Khandan
  • 11
  • 2
  • What would it mean for a particular variable to be "more crucial" & another to be 'less crucial' in this context? – gung - Reinstate Monica Jun 25 '16 at 23:09
  • I have 25 variables and I want to reduce the dimension of may variables and remove those variables which are highly correlated and say the same thing – Khandan Jun 26 '16 at 08:44
  • There are just [four other questions](http://stats.stackexchange.com/questions/tagged/dimensionality-reduction+time-series) tagged [tag:dimensionality-reduction] and [tag:time-series], but perhaps you could check [tag:dimensionality-reduction] on its own. [This question](http://stats.stackexchange.com/questions/82291/time-series-dimensionality-reduction) has an interesting answer, in my opinion, but perhaps it is not very relevant to your case. – Richard Hardy Jun 26 '16 at 11:38
  • @Khandan As your observations always contain all variables: would looking at each observation individually + using simple feature correlation over all such observations not be an option for you? – geekoverdose Jun 27 '16 at 15:26
  • Dear gung I want to use these variable for forecasting but if I want to use all of them it will cause a high computation cost so I should reduce the number of variables. – Khandan Oct 28 '16 at 06:28
  • Add the @ symbol right in front the user name, then the user will get a notification. The way you wrote now gung will not be notified and will likely miss your comment. – Richard Hardy Oct 29 '16 at 15:52

0 Answers0