Fig. 3

Coefficients of linear regression models for peak year disruption (circles, bright colours) and impact (crosses, dim colours) in Computer Science (top row) and Physics (bottom row). Each column displays results obtained with a different key variable of our analysis, i.e., relative effort (left), relative productivity (center), relative time devoted (right). Magenta (blue) symbols refer to the coefficients associated with key regressors (control variables) in each model. Key regressors are highlighted in the plot’s titles. Error bars represent standard errors multiplied by three. From left to right, in Computer Science models for disruption have an \(R^2\) of 0.142, 0.132 and 0.116, while models for impact have an \(R^2\) of 0.009, 0.007 and 0.011. In Physics, models for disruption have an \(R^2\) of 0.086, 0.078 and 0.071, while models for impact have an \(R^2\) of 0.001, 0.001 and 0.002.