Skip to main content
. 2020 May 26;57(3):899–925. doi: 10.1007/s13524-020-00886-4

Fig. 1.

Fig. 1

Bias in estimated relationship between an individual-level characteristic η and odds of abortion (expressed in log odds), according to overall abortion prevalence and degree of differential reporting by characteristic η, estimated using Monte Carlo simulations. True log odds (β1) are fixed at 1.