Table 1.
Parameters used for data generation.
Table 2.
Median (Interquartile range (IQR)) absolute percent biasa and mean squared error (MSE) for the regression coefficient as estimated via QUAD or PQL, overall and by data generation parameters.
Table 3.
Median (Interquartile range) absolute percent biasa and mean squared error σ2u as estimated via QUAD or PQL, overall and by data generation parameters.
Figure 1.
Boxplot depicting the slopes from separate simple linear regressions for the effect of the absolute percent difference in ORPQL and ORQUAD on the absolute percent bias in ORQUAD or ORPQL, respectively, overall and by data generation parameters.
Median (interquartile range) of the estimated slope is the center of the box, box edges are the 25th and 75th percentile respectively, ends of the dashed lines are the 10th and 90th percentile, respectively.
Figure 2.
Barplot depicting the proportion of scenarios in which the effect of the absolute percent difference in ORPQL and ORQUAD was a statistically significant predictor of the absolute percent bias in ORQUAD or ORPQL, respectively from separate simple linear regressions, overall and by data generation parameters.
Figure 3.
Boxplot depicting the R2 from separate simple linear regressions for the effect of the absolute percent difference in ORPQL and ORQUAD on the absolute percent bias in ORQUAD or ORPQL, respectively, overall and by data generation parameters.
Median (interquartile range) of the R2 is the center of the box, box edges are the 25th and 75th percentile respectively, ends of the dashed lines are the 10th and 90th percentile, respectively.
Table 4.
Results from a linear mixed quantile regression model with absolute percent bias in the odds ratio estimated via PQL or QUAD as the dependent variable and absolute percent difference in the odds ratios as estimated via PQL and QUAD as the independent variable, adjusted for data set characteristics (β1, σ2u, proportion with the outcome (p), total number of observations in the data set and data set composition).
Figure 4.
Boxplot depicting the slopes from separate simple linear regressions for the effect of the absolute percent difference in σPQL and σQUAD on the absolute percent bias in σQUAD or σPQL, respectively, overall and by data generation parameters.
Median (interquartile range) of the estimated slope is the center of the box, box edges are the 25th and 75th percentile respectively, ends of the dashed lines are the 10th and 90th percentile, respectively.
Figure 5.
Barplot depicting the proportion of scenarios in which the effect of the absolute percent difference in σPQL and σQUAD was a statistically significant predictor on the absolute percent bias in σQUAD or σPQL, respectively from separate simple linear regressions, overall and by data generation parameters.
Figure 6.
Boxplot depicting the R2 from separate simple linear regressions for the effect of the absolute percent difference in σPQL and σQUAD on the absolute percent bias in σQUAD or σPQL, respectively, overall and by data generation parameters.
Median (interquartile range) of the R2 is the center of the box, box edges are the 25th and 75th percentile respectively, ends of the dashed lines are the 10th and 90th percentile, respectively.
Table 5.
Results from a linear mixed quantile regression model with absolute percent bias in σ2u estimated via PQL or QUAD as the dependent variable and absolute percent different in σ2u as estimated via PQL and QUAD as the independent variable, adjusted for data set characteristics (β1, σ2u, proportion with the outcome (p), total number of observations in the data set and data set composition).