Valid and precise modeling of self-reported problems with condom use is necessary for research on condom safety and effectiveness. Traditional techniques may produce misleading estimates when modeling large proportion of zero values (perfect users or non-reporters) and discrete nonzero counts (at risk of reporting problems). Zero-inflated Poisson mixed regression allowed us to identify the characteristics of the two groups: that older women were significantly more likely to be non-reporters or perfect users. The nonzero problem rate decreased during follow-up, and was lower among women who believed in the benefits of condom use, and had no sexually transmitted diseases at baseline.
Length of hospital stay, and number of surgeries for children presenting with certain conditions, offer examples of excess numbers of zeros in pediatric research setting. Zero-inflated regression techniques can produce the mixture of correct probability distributions, examine the effects of study variables, and conclusively draw accurate inferences about child health.