epistasis odds ratio confidence intervals

[Guostė Petrauskaitė]

Hi all!

I wish to know is it possible to calculate the confidence intervals for odds ratio for interaction (OR_INT) in epistasis analysis with PLINK. I tried with --ci 0.95 but it is not working. 

Thanks in advance!

[Christopher Chang]

No, this isn't directly supported by PLINK 1.x (though it's a planned optional column set for PLINK 2.0 --epistasis).

For now, you can infer this from the relation between the beta/odds-ratio and the chi-square statistic:

  BETA_INT = ln(OR_INT)

  chi-square STAT = (BETA_INT / SE)^2

  -> SE = BETA_INT / sqrt(STAT)

  95% confidence interval for BETA_INT is [BETA_INT - 1.96 * SE, BETA_INT + 1.96 * SE]; exponentiate these values to get the OR_INT confidence interval