Keywords: Jackknife, Empirical Likelihood, Medical Cost, Censored Regression
Recent studies show that medical cost data is heavily censored and highly skewed, which lead to applying parametric modeling becomes more complex and less accurate. In this paper, we propose to use empirical likelihood methods based on influence function and jackknife techniques to construct confidence region for vectors of median medical cost with regression parameters when data is heavily censored and skewed, and we will construct confidence intervals for the expected median costs. Simulation studies are conducted to compare coverage probabilities generated by proposed confidence region with the existing normal approximation-based confidence region. The proposed Jackknife Empirical Likelihood method are observed to have better performances than existing methods for finite sample and with large percentage of censoring. The new methods will also be illustrated through Canadian implantable defibrillator study cases.