Skip to main content

Table 3 Threshold criteria used in the studies and description of epistatic analyses.

From: Comparison of analyses of the QTLMAS XII common dataset. II: genome-wide association and fine mapping

Study Threshold
criteria
P-value Comments
LABayes 2× ln(Bayes factor) ≥ 3 0.08 We equated 2× ln(Bayes factor) with a likelihood
ratio test statistic with one degree of freedom [17]
and calculated the p-value with a χ2 approximation.
This method compared models with an increasing number of QTL,
for each chromosome, therefore far fewer tests were
necessary than for the other studies.
LDBayes - - No significance tests were performed.
LDLA1 LD: F > 4a 0.007 Tests were only performed on markers that had
a significant difference in allele frequency
between high/low offspring from each sire at p < 0.0016.
  LDLA: LRTb > 12.8 0.0003c Tests were only performed on markers that had
a significant difference in allele frequency
between high/low offspring from each sire
at p < 0.0016.
LDmulti   8×10-6 An epistatic analysis was performed.
An epistatic model was tested against two-locus
marginal model (p < 8 × 10-6)
for pairs of markers significant alone,
against one locus model (p < 1 × 10-7)
for one significant and one non-significant marker,
and null model for pairs of non-significant markers
(p < 3 × 10-9).
In last two cases, epistatic model was then tested
against two-locus marginal model at
p < 2 × 10-5 and p < 5 × 10-7, respectively.
LDHap haplotype: HQd ≥ 15 2×10-9e  
  single marker: LRT >32.8 10-8 An epistatic analysis was performed
using single marker association on pre-corrected data.
For each 1 cM interval the marker in highest
average LD with the others in the same interval was found.
Each type of epistatic interaction
(e.g. additive by dominance) was then tested
for pairs of these markers at p < 10-3.
When these were significant epistasis was tested
for all pairs of markers in the two intervals at
p < 10-6 and pairs within 10 cM of each other were excluded.
LDLA2 LRT > 6 0.014c -
  1. aLowest F-to-enter value reported. bLikelihood ratio test statistic. cFrom a χ2 approximation with one degree of freedom. dHannan-Quinn criteria, which is similar to 2× ln(Bayes factor). eFrom regression of p-values obtained by permutation test (108 replicates) for chromosome 1 with raw data against HQ score (Ledur, pers. comm.).