Limits...
Multi-level block permutation.

Winkler AM, Webster MA, Vidaurre D, Nichols TE, Smith SM - Neuroimage (2015)

Bottom Line: In a previous study, we defined exchangeability for blocks of data, as opposed to each datum individually, then allowing permutations to happen within block, or the blocks as a whole to be permuted.Here we extend that notion to allow blocks to be nested, in a hierarchical, multi-level definition.The strategy is compatible with heteroscedasticity and variance groups, and can be used with permutations, sign flippings, or both combined.

View Article: PubMed Central - PubMed

Affiliation: Oxford Centre for Functional MRI of the Brain, University of Oxford, Oxford, UK. Electronic address: winkler@fmrib.ox.ac.uk.

Show MeSH

Related in: MedlinePlus

Maps showing the locations of the peaks of significance, for positive (+) and negative (−) correlations of height, weight, and bmi with cortical surface area and thickness. For conciseness, and given their lack of overlap, the original maps for thickness were thresholded at 0.05 and added together, allowing the regions to be displayed in the same figure. Even after using fwer-correction across the brain and contrasts, the unrestricted shuffling identified seemingly significant regions; these regions were not found significant using the restricted permutations that respect the family structure in the hcp sample. Provided that these traits are highly non-independent between subjects (i.e., heritable) this suggests that these results, produced with simple, unrestricted permutation, are in fact false positives (the peaks of significance for both restricted and unrestricted are listed in Supplementary Table 3).
© Copyright Policy - CC BY
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4644991&req=5

f0050: Maps showing the locations of the peaks of significance, for positive (+) and negative (−) correlations of height, weight, and bmi with cortical surface area and thickness. For conciseness, and given their lack of overlap, the original maps for thickness were thresholded at 0.05 and added together, allowing the regions to be displayed in the same figure. Even after using fwer-correction across the brain and contrasts, the unrestricted shuffling identified seemingly significant regions; these regions were not found significant using the restricted permutations that respect the family structure in the hcp sample. Provided that these traits are highly non-independent between subjects (i.e., heritable) this suggests that these results, produced with simple, unrestricted permutation, are in fact false positives (the peaks of significance for both restricted and unrestricted are listed in Supplementary Table 3).

Mentions: Permuting the data freely to test the hypotheses of correlation between thickness or area and indices of body size, therefore not respecting the structure of the sibships, allowed the identification of a few seemingly significant associations, even after fwer correction across the whole brain, and considering that both positive and negative tests were being performed. These regions, shown in Fig. 10, are (1) the left anterior cingulate for a positive correlation between height and cortical surface area, (2) the right orbitofrontal medial cortex for a positive correlation between thickness and bmi, (3), the right temporal pole, at the confluence of the inferior temporal gyrus, for a negative correlation between thickness and body weight, and (4) the right inferior temporal gyrus for a negative correlation between thickness and height. All these regions are very small, two of them comprising just one vertex at the resolution of the surfaces. However, using the proposed multi-level permutation strategy, in which shufflings only happen within siblings of the same type, and in which families with identical structure are allowed to be permuted as a whole, therefore respecting the kinship structure, all these findings became no longer significant. Supplementary Table 3 shows the minimum (most significant) p-value throughout the brain for both unrestricted and restricted permutation.


Multi-level block permutation.

Winkler AM, Webster MA, Vidaurre D, Nichols TE, Smith SM - Neuroimage (2015)

Maps showing the locations of the peaks of significance, for positive (+) and negative (−) correlations of height, weight, and bmi with cortical surface area and thickness. For conciseness, and given their lack of overlap, the original maps for thickness were thresholded at 0.05 and added together, allowing the regions to be displayed in the same figure. Even after using fwer-correction across the brain and contrasts, the unrestricted shuffling identified seemingly significant regions; these regions were not found significant using the restricted permutations that respect the family structure in the hcp sample. Provided that these traits are highly non-independent between subjects (i.e., heritable) this suggests that these results, produced with simple, unrestricted permutation, are in fact false positives (the peaks of significance for both restricted and unrestricted are listed in Supplementary Table 3).
© Copyright Policy - CC BY
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4644991&req=5

f0050: Maps showing the locations of the peaks of significance, for positive (+) and negative (−) correlations of height, weight, and bmi with cortical surface area and thickness. For conciseness, and given their lack of overlap, the original maps for thickness were thresholded at 0.05 and added together, allowing the regions to be displayed in the same figure. Even after using fwer-correction across the brain and contrasts, the unrestricted shuffling identified seemingly significant regions; these regions were not found significant using the restricted permutations that respect the family structure in the hcp sample. Provided that these traits are highly non-independent between subjects (i.e., heritable) this suggests that these results, produced with simple, unrestricted permutation, are in fact false positives (the peaks of significance for both restricted and unrestricted are listed in Supplementary Table 3).
Mentions: Permuting the data freely to test the hypotheses of correlation between thickness or area and indices of body size, therefore not respecting the structure of the sibships, allowed the identification of a few seemingly significant associations, even after fwer correction across the whole brain, and considering that both positive and negative tests were being performed. These regions, shown in Fig. 10, are (1) the left anterior cingulate for a positive correlation between height and cortical surface area, (2) the right orbitofrontal medial cortex for a positive correlation between thickness and bmi, (3), the right temporal pole, at the confluence of the inferior temporal gyrus, for a negative correlation between thickness and body weight, and (4) the right inferior temporal gyrus for a negative correlation between thickness and height. All these regions are very small, two of them comprising just one vertex at the resolution of the surfaces. However, using the proposed multi-level permutation strategy, in which shufflings only happen within siblings of the same type, and in which families with identical structure are allowed to be permuted as a whole, therefore respecting the kinship structure, all these findings became no longer significant. Supplementary Table 3 shows the minimum (most significant) p-value throughout the brain for both unrestricted and restricted permutation.

Bottom Line: In a previous study, we defined exchangeability for blocks of data, as opposed to each datum individually, then allowing permutations to happen within block, or the blocks as a whole to be permuted.Here we extend that notion to allow blocks to be nested, in a hierarchical, multi-level definition.The strategy is compatible with heteroscedasticity and variance groups, and can be used with permutations, sign flippings, or both combined.

View Article: PubMed Central - PubMed

Affiliation: Oxford Centre for Functional MRI of the Brain, University of Oxford, Oxford, UK. Electronic address: winkler@fmrib.ox.ac.uk.

Show MeSH
Related in: MedlinePlus