diff --git a/slides/sections/9.md b/slides/sections/9.md index e7487fd..4d112f1 100644 --- a/slides/sections/9.md +++ b/slides/sections/9.md @@ -3,14 +3,17 @@ ## Summary ::: incremental -- 5 out of 6 tests properly work for a high number of points ($N = 50000$) + +- All tests work reliably for large samples ($N \geq 5000$) + - Properties comparison: - - Median estimation works; decreases in significance as $N$ decreases - - HSM swtill properly works for $N = 1000$ - - KDE is the least accurate test (not working for Landau distribution) -- KS still properly works for very few points ($N = 200$) -- Trapani test, although less informative, till works for $N = 1000$ - (not working for $N = 200$) + - Median significance correlates well with $N$ + - HSM still significant at small sample size ($N = 1000$) + - FWHM unreliable at moderate sample size ($N \leq 5000$) + +- KS accurate even at very small sample ($N = 200$) + +- Trapani still works for $N = 1000$ but fails for smaller sizes :::