Rutkowski ([2011]) pointed out that the estimation of subgroup differences can be problematic if the proportions of missing values on the background variables differ between groups.
This is particularly problematic where significant proportions of health care are delivered outside of the health care system.
Misclassification of chronically-infected individuals as recently infected is particularly problematic, since the proportion of individuals with chronic HIV infection in a population often greatly exceeds the proportion of individuals with recent HIV infection [12].
Furthermore, this over-calling of significant interactions becomes more problematic as the proportion of duplicate di-tags increases.
Patellofemoral pain syndrome (PFP) is a common musculoskeletal condition that has a tendency to become chronic and problematic in a proportion of affected individuals.
For all models, individuals classified into more problematic classes showed higher proportions of early sexual intercourse, infrequent condom use, receiving money for sexual services, and ever contracting an STD.
For example, North American data suggest that only limited proportions of problematic drug users receive basic social or health services [ 41- 43], and receive targeted interventions or services for key health risks, such as BBV/STD risks or problems, only in exceptional cases [ 44].
HTLV-1/2 diagnosis in high-risk populations from São Paulo, Brazil has been problematic due a high proportion of seroindeterminate results.
Another reason is that, if the proportion of problematic UEs in the cell is small, the abnormal values are hidden when the average over all the UEs in the cells is calculated.
In empirical applications where background variables typically are correlated, it is plausible to assume that the more background variables are included, the less problematic is a high proportion of missing values on a single variable as other variables might compensate the loss of information.
