In this project we aim to take one step further the work on preferential sampling in a geostatistical setting by addressing preferential sampling in the longitudinal setting, as many times not only the observed longitudinal measurements give information on a disease state but also the follow-up times help to understand the medical condition of the patients. We intend to apply a model tested through simulations studies in the previous year, to a data set of an observational study on monitoring kidney function. The analysis will be compared with the classic longitudinal analysis. We conclude in previous simulation studies that that ignoring preferential designs can originate seriously misleading inferences. |