During the early stages of statistical analysis and modeling, analysts often perform graphical exploratory data analysis (EDA) to identify characteristics of the data and inform the following steps. However, different statistical graphics can highlight distinct features of the same data and induce analysts to make different modeling choices. We conducted two experiments with undergraduate statistics students. The first experiment assessed whether graphic choices influenced students’ ability to identify non-null data and whether the features of the data distribution they relied on when making decisions varied with graph types. In a second experiment, we assess whether different graphical presentations are associated with different modeling decisions. We found that different graphical representations of the same data highlight distinct characteristics of the distribution and can result in differential identification of non-null data. We describe associations between graph types and modeling decisions. We also assess variability in modeling responses, which appears to be reduced by additional education.