Figure 1.
Summary of the steps involved in a FAM-MDR analysis.
The figure shows the three steps of FAM-MDR Part II on one of the simulated datasets for Model M27, ,
and
, for the analysis without main effects correction.
Table 1.
Relative importance of main effects and epistatic variances for different two-locus models.
Figure 2.
Probability-probability plots based on 100 replicates under the complete null hypothesis of no association.
Data are generated with no main effects and no two-way interaction effects, with and
. Analyses are performed without correction for main effects. The first three panels show results of FAM-MDR (A), FAM-MDR* (B) and PGMDR (C) for the usual situation study considering 10 SNPs. The final panel (D) shows PGMDR results when only 2 SNPs are considered. The straight lines indicate the theoretical probability-probability curve (light blue) and the 5% significance level (dark blue).
Table 2.
Type I error rates under the general null hypothesis of no association.
Table 3.
Type I error rates of FAM-MDR under the null hypothesis of no epistasis.
Figure 3.
Power of FAM-MDR and PGMDR based on 100 replicates, with .
The different panels show results for M27 (A) and M170 (B). Abbreviations: Corr. = with main effects correction, No Corr. = without main effects correction.
Table 4.
Power of FAM-MDR, FAM-MDR* and PGMDR.
Table 5.
Epistasis analyses of Amish Family Diabetes Study data.