Re: Variance estimation by Permutation Samples

Follow-up

I made 20 experiments, each one using a different source-sample from N(100, 20):20, involving the Tests 1, 2, and 3.

Test 1 - 19 successes, 1 fail(*)

Sample #19 , Sample variance = 180.6
Test 1 _________[104.48 ; 385.34*]
Test 2 __________[99 ; 258*]
Test 3 __________[51 ; 329*]

Test 2 - 3 fails

Sample #19

Sample # 9 , Sample variance = 270.9
Test 1 _________[156.70 ; 577.91]
Test 2 _________[195 ; 339*]
Test 3 _________[ 136; 386*]

Sample # 12 , Sample variance = 228.4
Test 1 _________[132.09 ; 487.16]
Test 2 _________[156 ; 297*]
Test 3 _________[105 ; 352*]

Data main features:

I - If the sample variance is 300 or less all three tests fail to include the true variance 400.
II - The Modified Moments test provides the shorter intervals, next the Bootstrap, the wide ones are those from the Parametric test.
III – The resampling tests (2 and 3) can fail where Test 1 succeed.


___________licas (Luis A. Afonso)
.

Quantcast