|
Abstract:
|
For typical longitudinal studies, the response and any time-varying covariate(s) are recorded at the same time points within each individual. In the case that these are measured at distinct time points, a pre-processing step is required before applying traditional modeling. We have thus proposed a "binning" approach where we divide the original data into equally spaced bins of time. To evaluate the effects of binning, we conduct here an extensive simulation study with a longitudinal binary response and a single continuous time-varying covariate, measured at distinct time points, consistent with a real data application of ours. We would like to provide guidelines regarding when the binning method will/will not be successful. Influential factors include within-subject variability, distance in time between when the response and the time-varying covariate(s) are measured, bin size, etc.
|
