Replication Materials for Leveraging a surrogate outcome to improve inference on a partially missing target outcome.
By: Zachary R. McCaw
Updated: 2021-12-04
Ensure the dependencies under scripts/dependencies.R are installed.
size_sim.sh provides a working example of simulations under the null hypothesis that genotype has no effect on the target outcome. Running sh size_sim.sh as is will perform 1000 simulations, at each of 16 different configurations, with a sample size of 100, tabulate the results, then construct a uniform qq plot for the p-values. The code may be scaled up by setting --testing FALSE.
power_sim.sh provides a working example of simulations under the alternative hypothesis, wherein genotype has an effect on the target outcome. Running sh power_sim.sh as is will perform 100 simulations at each of 160 different configurations, with a sample size of 100, tabulate the results, then construct power curves. The code may be scaled up by setting --testing FALSE.
eQTL summary statistics for SNP-transcript pairs with p-values <= 1e-3 are located under data. In each case the target tissue is the substantia nigra. The surrogate tissues include whole blood, skeletal muscle, and cerebellum. data/data_description.tsv provides a description of the columns in each of the summary statistic files. Correlations between the target and surrogate tissue, by transcript, among complete-cases, are provided under correlations.