Table 2.
Summary of simulated datasets
| Number of datasets | 10 |
| Number of genes per dataset | ∼ 15,000 |
| Sample sizes | 3, 5, and 10 per group |
| Number of observation per subject | 4 |
| Model Parameters | |
| βg1: Difference in log(expression) between treatment and control at baseline | 0 (all genes) |
| βg2,βg3,βg4: Change in log(expression) over time in the control group | 0 (all genes) |
| βg5,βg6,βg7: Difference in change in log(expression) over time between the treatment and control groups | 0 (80% of genes), βg5=±1/3,βg6=±2/3,βg7=±1 (20% of genes) |
| Drawn from an empirical distribution based on human samples in real RNA-seq data sets with repeated measures [36, 37] | |
| Significance tests | |
| Between-subject | Are there differences in expression between the treatment and control at any of the time points? H0:βg1=βg1+βg5=βg1+βg6=βg1+βg7=0 |
| Within-subject | Is there a change in gene expression between any timepoints for the treatment group? H0:βg2+βg5=βg3+βg6=βg4+βg7=(βg2+βg5)−(βg3+βg6)=(βg2+βg5)−(βg4+βg7)=(βg3+βg6)−(βg4+βg7)=0 |
| Interaction | Are there any significant interaction effects? βg5=βg6=βg7=0 |
| Global | Are there any significant model coefficients? H0:βg1=βg2=βg3=βg4=βg5=βg6=βg7=0 |