The positive association between exposure to sexual content and pregnancy became even stronger when all covariates were included in the multivariate regression model ( .44; P .03) (Table 2). To illustrate the magnitude of this effect, we used the full sample to translate the coefficients from our multivariate regression model into predicted probabilities of pregnancy at the 10th (low), 50th (medium), and 90th (high) percentiles of baseline exposure to sexual content on television, holding all other covariates at their naturally occurring values.