Barely Significant

Excerpts

33,758 sentences · 1 filter applied

Because only published and nominally significant values are considered, this makes effect estimates arguably insensitive to non-retrievable unpublished (and therefore likely non-significant) results and makes it possible to assess more insidious forms of dissemination bias such as p -hacking (e.g., repeated calculation of significance statistics in primary studies by means of different methods or inclusion of different a posteriori-selected covariates, until nominal significance is achieved).
Results We found lower HIV/STI incidence density in the intervention relative to the control group in 5 sites we labeled as “program effective sites,” but not in 8 sites we labeled as “program ineffective sites.” Using generalized estimating equations controlling for site and computed mean difference effect sizes, we examined statistically and practically significant differences, respectively, between the two groups of sites along various client-, provider-, organization-, and structure-related characteristics.