Table 1.

Summary of asymptotic biases in genotype-genotype interactions under various population stratification conditions for two ethnicities

Comparison of high-risk and low-risk ethnicities
Bias in estimation
Figure
Baseline disease risksMarginal genotype distributions for gene 1 (G1)Marginal genotype distributions for gene 2 (G2)Linkage disequilibrium (D) between G1 and G2G1 main effectG2 main effectInteraction between G1 and G2
=== (condition a)±±+/−*
≠(condition b)+++1
= (D = 0; condition c)±++/−2
≠(condition d)+++
== (D = 0; symmetry condition c)+±+/−
≠(cf. condition d)+++
= (D = 0; condition e)+++3
≠(condition f)+++
=Any combination of conditions
  • NOTE: ≠, variables/distributions were different across the high-risk and low-risk ethnicities. =, variables/distributions were same across the high-risk and low-risk ethnicities. ±, bias was generally negligible, towards the null when the effect was nonzero; bias was absent when the effect was zero. +/−, bias was generally negligible; bias was absent when both main effects and interactions were all zero. +, bias was present. −, no bias.

  • * No figure was presented because bias (if any) was not due to population stratification and was generally negligible.

  • No figure was presented because (by symmetry of G1 and G2) similar results were presented in Fig. 2.

  • No figure was presented because no bias was observed.