Estimation of Survival with a Combination of Prevalent and Incident Cases in the Presence of Length Bias