Fig. 4

Modeling framework for the estimation of ovulation and menstrual states. a Modeling framework for the estimation of ovulation timing. (Top) Schematics of the 10-states HMM which discretizes the menstrual hormonal events (HM, heavy menses; LM, light menses; LE, low estrogen; HE, high estrogen; Ovu, ovulation; Rise, progesterone/BTT Rise; HP, high progesterone; EP, estrogen peak in luteal phase; LP, low progesterone). Arrows indicate possible state-transition; arrow thickness is not representative of actual transition probabilities (Methods). (Bottom) Examples of menstrual state estimation for the 2rd and 3rd cycle of 2 users. (Top of each chart) Original user observations as in Fig. 2a. (Middle of each chart) Colored squares HMM-labeled line) represent the most likely sequence of HMM states given the observations (Methods). (Bottom of each chart) Normalized probabilities of each state on each day of the cycle (Methods). b (Top) Cycle length and estimated ovulation day. (Bottom) Luteal phase duration, computed as the number of days between the ovulation day (excluded) and the 1st day of the next cycle (excluded). Vertical lines indicate median values. 80,708 (K) +24,119 (S) cycles with reliable ovulation estimation were used (Methods). c Average estimated state probabilities by cycle-day counting from estimated ovulation aggregated by total cycle length (in bins of 3 units) for all cycles with reliable ovulation estimation