We would like to submit this erratum to our paper titled “Leveraging Prognostic Baseline Variables to Gain Precision in Randomized Trials” by Colantuoni and Rosenblum, Statistics in Medicine, 2015, 34(18), 2602–2617. There was an error in the SAS code that was linked as supplementary material to our manuscript. The error involved the following lines of code: 248–9, 255–6, 263–4, 270–1. This error does not affect any of the results from the paper or supplementary material since these results were generated based on the R code (also provided in the supplementary material). The SAS code was provided in order to make the methods in the paper available to a wider audience. We apologize for our error in the SAS code, which we have now corrected. We made an additional improvement to both the R and SAS codes, in which users now have the option to truncate and standardize the weights involved in computing the PLEASE estimator. Truncation allows the user to limit the influence of participants with corresponding large weights. The standardization is done to improve the convergence in the corresponding iterative reweighted least squares computations. The revised R and SAS codes are located at http://people.csail.mit.edu/mrosenblum/papers/PLEASEcode.zip for downloading. We would like to thank Asanao Shimokawa and Antoine Chambaz for bringing the SAS coding error to our attention.
ASJC Scopus subject areas
- Statistics and Probability