Confounding adjustment performance of ordinal analysis methods in stroke studies
Fig 1
Bias of Mann-Whitney measure of superiority (MWS), ordinal logistic regression (OLR), stratified Mann-Whitney measure of superiority (sMWS), adjusted ordinal logistic regression (aOLR), and propensity score based stratified Mann-Whitney measure of superiority (psMWS) in scenario 1.
The psMWS was not performed in the scenario with one confounding variable. Runs (N): 1000. The x-axis shows the number of confounding variables modeled. The y-axis shows the bias, with estimates from the OLR analyses converted to a MWS. Boxplots display the lower adjacent value, 25th percentile, median, 75th percentile, and upper adjacent value (extreme outliers are not displayed).