Download presentation
Presentation is loading. Please wait.
Published byHarold Skinner Modified over 6 years ago
1
A nearly optimal sequential testing approach to permutation-based association testing Julian Hecker, Ingo Ruczinski, Brent Coull, Christoph Lange Introduction Permutation-based approaches address statistical issues in genetic association studies as: small sample sizes, rare genetic variants, design imbalances or non-normal phenotypes Significance levels of interest are usually very small ๏ permutation-based approaches are computational intensive Adaptive permutation strategies in existing genetic association analysis tools (e.g. PLINK Chang et al. 2015) use heuristics to stop early if the p-value is obviously non-significant Sequential testing approach ๐ true, unknown association p-value sequence ๐ฅ 1 , ๐ฅ 2 ,โฆ where ๐ฅ 1 =1 iff permuted statistic more extreme, 0 otherwise We introduce a small indifference region and consider the hypotheses ๐ป 1 :๐โค ๐ 1 and ๐ป 2 :๐โฅ ๐ 2 = ๐ 1 +๐ (e.g. ๐ 1 =5โ 10 โ8 and ๐= 10 โ8 )
2
A nearly optimal sequential testing approach to permutation-based association testing Julian Hecker, Ingo Ruczinski, Brent Coull, Christoph Lange Objects Pre-specified error probabilities ๐ผ 1 , ๐ผ 2 (e.g. ๐ผ 1 = ๐ผ 2 = 10 โ10 ). Define (Pavlov 1991) ๐ ๐ ๐ผ ๐ โmin{๐: ๐ ๐ / sup ๐โ ๐ท ๐ ๐ ๐ ๐, ๐ฅ ๐ โฅ ๐ผ ๐ โ1 } for ๐=1,2, where ๐ท 1 = 0, ๐ 1 , ๐ท 2 = ๐ 2 ,1 , ๐ฅ ๐ = ๐ฅ 1 ,โฆ, ๐ฅ ๐ , ๐ ๐ ๐, ๐ฅ ๐ = ๐=1 ๐ ๐( ๐, ๐ฅ ๐ ), ๐ ๐ โ ๐=1 ๐ ๐( ๐ ๐โ1 , ๐ฅ ๐ ) and ๐ฝ ๐โ๐ โ ๐=๐ ๐โ๐ ๐ ๐ + ๐ ๐ ๐ . ๐(๐,๐ฅ) Bernoulli density with parameter ๐. Decision procedure STr If ๐ 1 ๐ผ 1 โค ๐ 2 ( ๐ผ 2 ), we set ๐=2 and ๐= ๐ 1 ( ๐ผ 1 ). If ๐ 1 ๐ผ 1 > ๐ 2 ( ๐ผ 2 ), we set ๐=1 and ๐= ๐ 2 ( ๐ผ 2 ).
3
A nearly optimal sequential testing approach to permutation-based association testing Julian Hecker, Ingo Ruczinski, Brent Coull, Christoph Lange Theorem (Pavlov 1991, Tartakovsky 2014) 1.) ๐ ๐ ๐ฟ=2 โค ๐ผ 1 for ๐โ ๐ท 1 and ๐ ๐ ๐ฟ=1 โค ๐ผ 2 for ๐โ ๐ท 2 2.) Let ๐พ ๐ก 1 , ๐ก 2 ,๐ผ be the class of all decision rules ( ๐ โฒ , ๐ โฒ ) such that ๐ ๐ ๐ฟโฒ=2 โค ๐ก 1 ๐ผ for ๐โ ๐ท 1 and ๐ ๐ ๐ฟโฒ=1 โค ๐ก 2 ๐ผ for ๐โ ๐ท 2 , then ๐ธ ๐ ๐ inf ๐ โฒ , ๐ โฒ โ๐พ ๐ก 1 , ๐ก 2 ,๐ผ ๐ธ ๐ ๐ โฒ =1+๐ 1 as ๐ผโ0 for all ๐โ 0,1 . Error probabilities are strictly controlled Approaches theoretical minimum number of expected permutations if error level goes to zero
4
A nearly optimal sequential testing approach to permutation-based association testing Julian Hecker, Ingo Ruczinski, Brent Coull, Christoph Lange Comparison with confidence interval based approach ๐ empirical estimate of p-value after ๐ permutations ( ๐ โ ๐ ๐ผ ๐๐ธ, ๐ +๐ ๐ผ ๐๐ธ) corresponding 1โ๐ผ confidence interval CI-based rule (CIr) choose ๐=1 if ๐ + ๐ ๐ผ ๐๐ธโค ๐ 2 , set ๐=2 if ๐ โ ๐ ๐ผ ๐๐ธโฅ ๐ 1 Ratio STr/CIr ratio overall number of permutations needed for 12,045,191 single variant tests in LD (simulated) Red scenario: Type 1 error at least for CIr ๐ ๐ / ๐ ๐ ๐ถ ๐ / ๐ถ ๐ ๐ถ CIr ratio STr/CIr 1e-09/2e-09 1e-10/1e-10 1e-10 12.62 5e-08/6e-08 6.39 9e-04/1e-03 1e-10/4e-03 1.11
5
A nearly optimal sequential testing approach to permutation-based association testing Julian Hecker, Ingo Ruczinski, Brent Coull, Christoph Lange Our approachโฆ โฆ can be applied to arbitrary association test statistics. โฆ controls the error probabilities rigorously and approaches the theoretical minimum in terms of expected number of permutations. โฆ is easy to implement and no additional computational effort is required. โฆ is an alternative approach to sequential Monte Carlo hypothesis testing (Gandy 2009). Example code available at: Contact: Julian Hecker
Similar presentations
© 2025 SlidePlayer.com Inc.
All rights reserved.