The Elmira sample included two deviations from the randomization procedure. First, six housemates of women already randomly assigned and enrolled in the study were assigned to the same treatment as the women already enrolled. Second, the probability of being assigned to one of the treatment groups was increased in the last 6 months of the 30 month enrollment period. The first issue suggests a mismatch between the unit of assignment (adult in the household) and the unit of analysis, which may lead to overstating the precision of the standard errors. The second issue should lead to a weighting strategy in the analysis, so that those who were enrolled later receive less weight in the analysis. Weighting, however, was not used in these studies.
Eckenrode, J., Zielinski, D., Smith, E., Marcynyszyn, L. A., Henderson, C. R., Kitzman, H., et al. (2001). Child maltreatment and the early onset of problem behaviors: Can a program of nurse home visitation break the link? Development and Psychopathology, 13(4), 873–890.