A substantial body of literature has examined how women’s psychology and behavior vary as a function of conception risk across the ovarian cycle. These effects are widely believed to be outcomes of hormonal regulation, in particular effects of estrogens (E) and progesterone (P). Increasingly, researchers have sought to test predictions about how psychological processes or behavior vary as a function of conception risk by examining associations with estrogen (e.g., estradiol) and progesterone levels. Yet issues regarding how best to assess these associations arise. Should hormone levels be log-transformed? Do hormone ratios best capture their joint effects? How important are hormone interactions? How should outliers be treated? Across two large datasets, we examined hormonal predictors of conception risk, estimated from day of a luteinizing hormone (LH) surge. Log-transformed E and P levels predicted conception risk better than raw E and P levels did. The raw E/P ratio was a relatively poor predictor, whereas the log-transformed ratio (ln[E/P]) was a relatively good predictor. E × P interactions were detected but weak. Outliers were frequent, especially in distributions of raw hormone levels. Hormone measures predicted two psychological outcomes in these datasets—sexual desire and preferences for strength and muscularity—in parallel to how strongly they predicted conception risk. These results give rise to several recommendations regarding treatment of hormone measures and their use in analyses.