Incorporating ceiling effects during analysis of speech perception data from a paediatric cochlear implant cohort