A bootstrap approach for the reliable estimation of digestible starch fractions from digestograms

The estimation of confidence intervals for the rapidly (RDS) and slowly (SDS) digestible fractions is an important issue for assessing the nutritional characteristics of starch. Usually, the confidence interval estimation relies on a reduced number of experimental data, which may lead to unacceptable uncertainties, and hence to misleading conclusions. This work proposed a bootstrap approach for the estimation of confidence intervals for digestible starch fractions. Based on experimental digestograms, the approach obtains many (hundreds or thousands) bootstrapped digestograms by carrying out sampling with replacement of the experimental data. A simple first-order exponential model is then used for the least-squares fit of the bootstrap digestograms, obtaining a reliable estimation of the statistical uncertainty of the model parameters. The fitted model is used to estimate the RDS and SDS fraction by reading out the fitted curves at 20 and 120 min. The distribution of these starch fractions is used to estimate the 95% confidence intervals and to carry out ANOVA (p < 0.05) for the classification of starch groups (i.e., significant differences). The approach implementation is simple, relying on basic statistic concepts handled by most practitioners. The method was illustrated with an example of the dry heat treatment of corn starch.