From 2015 onwards, the DORA research program has used structural equation modeling (SEM), a predictive statistical modeling technique used to test relationships. We utilize PLS-PM (partial least squares path modeling) for our analysis, for several reasons: it does not require assumptions of normality in the data, it is will suited to exploratory and incremental research, and the analysis optimizes for predicction of the dependent variable (vs. testing for model fit of the data). All paths shown in the SEM figures are p < .05.

Below you can find the SEMs used in the 2021 analysis. Each box represents a construct we measured in our research, and each arrow represents relationships between the constructs. A larger box that contains boxes (constructs) is a second-order construct.

To interpret the model, all arrows can be read using the words *predicts, affects, drives,* or *impacts*.