In this paper, we present a comparative study of five methods for the estimation of missing values in building sensor data. The methods that were implemented and evaluated include linear regression, weighted K-nearest neighbors (kNN), support vector machines (SVM), mean imputation and replacing missing entries with zero. Using data collected from an actual office building, the methods were evaluated using varying parameter settings. Correlation based feature selection is used to evaluate how using different subsets of attributes may affect each method’s performance. We also evaluate the effect of including lagged variables as predictors. To test the robustness of each method, the amount of missing values were varied between 5% and 20%.