1

Problem: We have distance between user and a static device(in meters). We have lab tested data where user is actually standing away from device(actual values). Based on these we try to determine what is approximate distance of user(predicted interval range).

There are N confidence intervals. The sample data set used is same for all N algorithms.

ConfidenceInterval-A(95%) : Derived using Algorithm A.

ConfidenceInterval-B(95%) : Derived using Algorithm B.

. . .

ConfidenceInterval-N(95%) : Derived using Algorithm N.

We also have actual dataset values

I would like to certainly compare these Confidence Intervals and choose the best fit.

Questions:

  1. What factors should I consider when comparing two confidence intervals derived from same dataset but different algorithms ?

  2. What tests should I perform to confirm hypothesis later ?

TIA.

camelCase
  • 11
  • 1
  • Two properties can be used to rate the usefulness of a 9%% CI: (a) narrower is better than longer) (b) For each value of the parameter $\theta$ being tested, the 'coverage probability' of the CI should be 95% as claimed. // I guess it goes without saying that a CI for $\theta$ should not include impossible values of $\theta.$ // There are several competing CIs for the binomial success probability (see Wikipedia). If you want to explore in more detail, a paper (2001) in _Statistical Science,_ by Brown, Cai, & DasGupta) discusses the properties of some of them. – BruceET Feb 28 '22 at 23:33

0 Answers0