Currently, engineering research is highly data-oriented. Probabilistic methods to uncover patterns in noisy data are an essential part of this trend. The aim of this study was to demonstrate two of those concepts to the field of elastomer research and thermal aging: statistical hypothesis tests of experimental data and model selection. The two concepts are used to quantitatively assess whether the fracture stress of the rubber of interest only degrades due to short-term thermal oxidative aging, or if the fracture stress also improves for some aging temperatures. First, model-free statistical hypothesis tests are directly applied to fracture stress data. These tests have the advantage that no aging models are required, and hence, their results are unaffected by modeling choices. Their disadvantage is that not all experimental data can be considered. To circumvent this problem, a number of aging models are subjected to model selection to investigate whether growth of the fracture stress should be included in the aging model, or only degradation. However, model selection is also not perfect, because it only compares the models proposed by the practitioner. Because each method overcomes the shortcomings of the other, the combined application of the two methods makes a strong investigative tool that can be applied to any property affected by any physical phenomenon, for example, crack length due to ozone attack, modulus change due to thermal aging, and surface embrittlement due to photo-oxidation.