Fig 1.
Mean-Dispersion Plot with Fitted Dispersion Models.
The mean-dispersion plot with six fitted dispersion models (common, NBP, NBQ, trended, tagwise-common and tagwise-trend) for the Arabidopsis RNA-Seq dataset (19,623 genes from three biological samples in the mock treatment group). For better visualization, we use dots instead of curves for the tagwise procedures to indicate the variability of individual NB2 dispersion parameters about the trend.
Fig 2.
Empirical Probability Plots for Evaluating NB2 and NBP Model Fits.
Empirical probability plots with GOF test p-values for evaluating NB2 and NBP model fits on the earthquake dataset (sample size: 45), with 95% prediction envelope in dashed blue lines and 95% simultaneous prediction band in solid red lines (based on 999 Monte Carlo simulations). Points outside the prediction envelope are flagged as blue triangles.
Table 1.
Type-I Error Rate Evaluations.
Table 2.
Rejection Rate (Power) Evaluations.
Fig 3.
Empirical probability plots and GOF p-values for testing NB2 (top row) and NBP (bottom) on four simulated datasets with sample size = 45.
The simulated response distributions are (left to right): NB1, NB2, NB2 with outliers and NB2 with random (0, 22) noise added to log(ϕ). We superimpose 95% prediction envelopes in dashed blue lines and 95% simultaneous prediction bands in solid red lines (based on 999 Monte Carlo simulations). Points outside the prediction envelope but inside the simultaneous confidence bands are flagged as blue triangles, and points outside the simultaneous confidence bands are flagged as red crosses.
Fig 4.
Uniform QQ plots of individual GOF test p-values for the Arabidopsis dataset.
The results are based on a random sample of 1,000 genes from six experimental units in two experimental groups. The tagwise-common model (not shown) has a very similar pattern to the tagwise-trend model.