Fig. 4
From: A comparison of residual diagnosis tools for diagnosing regression models for count data
Performance of the RQRs in detecting covariate non-linearity effect of a sample dataset of size n=1000. The panels in the first row present the RQRs for the true fitted model: NB model with quadratic covariate effect (i.e., exp(β1x2)). The panels in the second row present the RQRs for the fitted wrong model: NB model with linearity covariate effect (i.e., exp(β1x)). The first two columns display the scatter plots and QQ plots of the RQRs, respectively. The red dashed lines in the QQ plots represent the simulated envelopes. The third column presents the histograms of the SW p-values for the RQRs over 5000 randomly generated datasets from the true model