depictr.influence_plot

depictr.influence_plot(model, title=None)

Bubble plot of studentised residuals against leverage.

Each point is one observation; the bubble area is proportional to its Cook’s distance, the standard measure of how much the fit would move if the point were dropped (Cook, 1977). Points to the right have high leverage (an unusual predictor combination), points far above or below zero are poorly fitted, and large bubbles are the ones that combine both into real influence.

Parameters:

• model (statsmodels results object) – A fitted OLS/GLM result.

• title (str, optional) – Plot title.

Return type:

plotnine.ggplot

References

Cook, R. D. (1977). Detection of influential observation in linear regression. Technometrics, 19(1), 15-18. https://doi.org/10.1080/00401706.1977.10489493