GWA Test Driver

Summary Report

Simulated Dataset #6

In these simulated data for 2000 individuals, we assumed 99% of genome-wide SNPs to be non-causal SNPs (i.e. under Null hypothesis with Odds Ratio=1) with their association p-values Uniformly distributed and remaining 1% SNPs to be causal SNPs with the Odds Ratio between 1.0 and 2.5 and distribution shape as Beta(0.5, 1.5).

Dynamic Power Plot and Table

This report tabulates and plots EDR estimates, both uncorrected and corrected for multiple testing, at pre-selected combinations of sample size and significance level.

The end user can request additional power records to be calculated and dynamically added to the table by filling out the text fields at the bottom of the table and clicking the submit button. This will add the user-specified power record to the table and update the plot.

family-wise significance level sample size EDR[1], uncorrected for multiple testing EDR, Bonferroni[2] EDR, FDR[3] EDR, mix-o-matic FP[4]
0.001 1000 0.133255 0.006328 0.028518 0.027073
0.001 2000 0.417457 0.1246 0.225766 0.221162
0.001 3000 0.68591 0.45093 0.551595 0.547688
0.001 4000 0.857297 0.771609 0.808865 0.807446
0.001 5000 0.94326 0.934276 0.936896 0.936762
0.01 1000 0.242636 0.011689 0.055024 0.053603
0.01 2000 0.531897 0.158684 0.2931 0.290062
0.01 3000 0.750775 0.489052 0.604158 0.601945
0.01 4000 0.881027 0.785923 0.827849 0.827053
0.01 5000 0.947755 0.935065 0.93897 0.938872
0.05 1000 0.367972 0.017934 0.088435 0.087105
0.05 2000 0.631161 0.187904 0.354184 0.352045
0.05 3000 0.802531 0.517924 0.646236 0.644832
0.05 4000 0.900692 0.796581 0.842986 0.84248
0.05 5000 0.9524 0.935827 0.941006 0.940933
 

NOTE: The runtime of an additional power record calculation depends on a number of factors, including the number of p-values in the dataset, the number of other users simultaneously requesting other calculations, etc. The expected runtime with no competition with other users is less than 1 minute per requested record.

NOTE: The '?' character in an EDR field indicates that the power calculation did not complete. See software specification for further detail and a description of situations where this might happen (e.g. during the calculation of the FDR-corrected significance level if there is little or no signal).

NOTE: For users interested in cutting and pasting the power table directly into a MS Excel spreadsheet, we have provided a demo video.

References

[1] Gadbury GL, Page GP, Edwards J, Kayo T, Prolla TA, Weindruch R, Permana PA, Mountz JD, Allison DB. Power and sample size estimation in high dimensional biology. Statistical Methods in Medical Research (2004) 13:325-338. DOI

[2] Bonferroni, C. E. 1936. Teoria statistica delle classi e calcolo delle probabilità. Publicazioni del R Istituto Superiore di Scienze Economiche e Commerciali di Firenze 8, 3-62.

[3] Benjamini, Y., and Hochberg, Y. (1995), Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing, Journal of the Royal Statistical Society, Ser. B, 57, 289-300. JSTOR

[4] Allison, D. B., Gadbury, G. L., Heo, M., Fern?ndez, J. R., Lee, C.-K., Prolla, T. A. and Weindruch, R. (2002). A mixture model approach for the analysis of microarray gene expression data. Comput. Statist. Data Anal. 39 1-20. DOI