Analysing outcome variables with floor effects due to censoring: A simulation study with longitudinal trial data