2.2.3. Significance tests
For each group of simulations described in Section 2.2.2, we averaged the outputs of simulations of all ensembles (each indi-vidual simulation of crop model forced by each climate model is considered a member of the ensemble) and created an average time-series (from 2011 to 2050) of soybean productivity, thereby reducing uncertainty and model-related bias. We then tested the hypothesis that the average soybean productivity changes from the first to the last decade in the 2011–2050 period due to cli-mate change. In other words, we test the hypothesis that soy
productivity in 2041–2050 (Y2041–2050 ) is different than the average soybean productivity in 2011–2020 (Y2011–2020 ), being that differ-ence related to the climate change that occurred between these
periods. We used the Student’s t-test, with a 5% level of signifi-cance and n = 10 years to test this hypothesis, in the two groups of simulations described in Section 2.2.2. We focus our discussion in statistically significant changes in soybean productivity until the middle of the century.