Skip to main content

Table 1 VIF and Type-I Error Control in the presence of a confounder

From: Quantitative gene set analysis generalized for repeated measures, confounder adjustment, and continuous covariates

    VIF Estimate Type-I Error
Sample Size Percent Conf. True VIF QuSAGE Q-Gen Qusage Q-Gen
5 0 20.024 19.04 18.89 .046 .054
5 .25 20.024 17.43 18.87 .042 .050
5 .5 20.024 17.89 18.89 .041 .055
5 0 6.435 6.20 6.16 .044 .050
5 .25 6.435 6.282 6.2 .040 .054
5 .5 6.435 8.04 6.16 .045 .055
5 0 1 0.99 0.99 .048 .053
5 .25 1 1.72 1.00 .041 .054
5 .5 1 4.16 0.99 .010 .057
15 0 20.024 19.75 19.74 .047 .047
15 .25 20.024 18.72 19.72 .056 .049
15 .5 20.024 18.59 19.71 .0822 .053
15 0 6.435 6.36 6.36 .053 .052
15 .25 6.435 6.43 6.36 .072 .052
15 .5 6.435 8.27 6.34 .101 .051
15 0 1 1.00 1.00 .048 .049
15 .25 1 1.719 1.00 .097 .050
15 .5 1 4.176 1.00 .074 .051
  1. VIF and type-I error estimates under a case-control simulation in the presence of a confounding variable where the percentage of genes that are affected by the confounder are examined for 0, 25, and 50 % respectively. Error rates in bold indicate they are within the margin of error (0.05 ± 0.00427) for the simulation study. Since Q-Gen allows for a linear model that adjusts for covariates, VIF estimation and the controlling of the type-I error is more consistent than the standard two sample t-testing conducted within QuSAGE