0

I am interested in how to create a (joint) confidence region (such as confidence ellipse) for a vector of parameters greater than 1.

It is easy to bootstrap many replications of the bootstrapped statistic by just returning a vector of length greater than 1 from the statistic function for the boot function in the boot package. But I don't know in R how to go from those replications to a, e.g., 95%, confidence region.

Note that I do not need to plot the region. I'm not sure how it would be represented internally in R. I would like to be able to check whether certain sets are a subset of the confidence region.

Xu Wang
  • 10,199
  • 6
  • 44
  • 78
  • 1
    See function `data.ellipse` in package `car`. – Rui Barradas Jun 21 '19 at 18:50
  • @RuiBarradas Thanks! That looks great (it seems in newer versions of car it is now called just `ellipse` and `data.ellipse` is deprecated). However, it seems those methods only work for two-dimensional parameters. I still cannot find any method that works for greater than 2 dimensions, not necessarily for plotting, just for checking whether they contain a subset or something like that. – Xu Wang Jun 22 '19 at 05:08

0 Answers0