Abstract:
|
In a cross-sectional study on time-to-pregnancy (TTP), study participants were asked to recall the duration from initiating attempts to conceive until conception occurred. Further they were asked about uncertainty attached to their response. Another challenge of this data is whether the pregnancy was planned or unplanned, with the TTP being well defined for planners. The objective is to estimate the TTP distribution from these complex survey data. We propose a multistage model with a model for binary outcome of planned pregnancy, as well as a model for TTP with a discrete survival function with random heterogeneity between subjects for time-to-pregnancy data and a model for the uncertainty indicator depending on time since pregnancy, through a multinomial regression. The ensuing likelihood contains different types of factor for the cases with different degrees of surety in her answer, which is conditioned on planned pregnancy. Our model allows for the consideration of covariate effects on each of the underlying factors of planners, measure of uncertainty and TTP distribution. We have additionally proposed prediction of survival distribution of TTP for a woman based on covariates.
|