A new way to estimate disease prevalence from random partial‐mouth samples