Improvements on a predictor–corrector strategy for parameter estimation with several data types