There is one point per observation.
The following show probability \(P_i\) on the \(x\)-axis:
\(P_i \times h_i\)Probability vs. leverage.
\(P_i \times \Delta P \chi^2_i\)Probability vs. the change in the standardized
Pearsons chi-squared
with observation \(i\) excluded.
\(P_i \times \Delta D_i\)Probability vs. the change in the standardized deviance
with observation \(i\) excluded.
\(P_i \times \Delta \hat{\beta}_i\)Probability vs. the change in the standardized
maximum likelihood estimators of the model coefficients
with observation \(i\) excluded.
\(P_i \times \Delta P \chi^2_i\)Bubbleplot of
probability vs. the change in the standardized
Pearsons chi-squared
with observation \(i\) excluded.
The area \(A_i\) of each circle is
proportional to \(\Delta \hat{\beta}_i\):
$$A_i = \pi r_i^2 \quad r_i = \sqrt{\frac{\Delta \hat{\beta}_i}{P_i}}$$
For details see:
?graphics::symbols
The following show leverage h_ih[i] on the x-axis:
\(h_i \times \Delta P \chi^2_i\)Leverage vs. the change in the standardized
Pearsons chi-squared
with observation \(i\) excluded.
\(h_i \times \Delta D_i\)Leverage vs. the change in the standardized deviance
with observation \(i\) excluded.
\(h_i \times \Delta \hat{\beta}_i\)Leverage vs. the change in the standardized
maximum likelihood estimators of the model coefficients
with observation \(i\) excluded.
The correlation of
\Delta \chi^2_i, \Delta D_i \mathrm{and} \hat{\beta}_i
dChisq, dDev and dBhat.
is shown in a pairs plot. See:
?graphics::pairs
The Value of dx is also returned, invisibly.