90 Language and legitimacy dynamics
We run our regressions by estimating robust standard errors and, to further
check the reliability of our results, we also run robust regressions using itera-
tively reweighted least squares. We find that our results hold also when running
robust regressions that use iteratively reweighted least squares.
Finally, we plotted graphs that represent the ‘leverage versus squared
residuals’ (Sall, 1990), to check for observations in our sample that show a high
leverage and a high residual, and thus may have an excessive influence on our
regressions. First, we plot the graph after the regression that estimates the eect
of SEVO vocabulary on sociopolitical legitimacy. In this case, the plot shows two
particularly pr ...