Fig. 2From: Inverse probability weighting to handle attrition in cohort studies: some guidance and a call for cautionMonte-Carlo variance obtained with CCA and the IPPW method (response models 1 and 10, see Table 2) for the nine response mechanism scenarios. The variance increased as the correlation between the exposure variable and the response variable increased for both methods. The variance was consistently higher with the IPPW method than with CCA in all scenarios. With the IPPW method, variance inflation was particularly observed when the exposure variable X was put into the response modelBack to article page