GLMResults.plot_ceres_residuals(focus_exog, frac=0.66, cond_means=None, ax=None) [source]

Produces a CERES (Conditional Expectation Partial Residuals) plot for a fitted regression model.

  • focus_exog (integer or string) – The column index of results.model.exog, or the variable name, indicating the variable whose role in the regression is to be assessed.
  • frac (float) – Lowess tuning parameter for the adjusted model used in the CERES analysis. Not used if cond_means is provided.
  • cond_means (array-like, optional) – If provided, the columns of this array span the space of the conditional means E[exog | focus exog], where exog ranges over some or all of the columns of exog (other than the focus exog).
  • ax (matplotlib.Axes instance, optional) – The axes on which to draw the plot. If not provided, a new axes instance is created.

fig – The figure on which the partial residual plot is drawn.

Return type:

matplotlib.Figure instance


RD Cook and R Croos-Dabrera (1998). Partial residual plots in generalized linear models. Journal of the American Statistical Association, 93:442.

RD Cook (1993). Partial residual plots. Technometrics 35:4.


cond_means is intended to capture the behavior of E[x1 | x2], where x2 is the focus exog and x1 are all the other exog variables. If all the conditional mean relationships are linear, it is sufficient to set cond_means equal to the focus exog. Alternatively, cond_means may consist of one or more columns containing functional transformations of the focus exog (e.g. x2^2) that are thought to capture E[x1 | x2].

If nothing is known or suspected about the form of E[x1 | x2], set cond_means to None, and it will be estimated by smoothing each non-focus exog against the focus exog. The values of frac control these lowess smooths.

If cond_means contains only the focus exog, the results are equivalent to a partial residual plot.

If the focus variable is believed to be independent of the other exog variables, cond_means can be set to an (empty) nx0 array.

© 2009–2012 Statsmodels Developers
© 2006–2008 Scipy Developers
© 2006 Jonathan E. Taylor
Licensed under the 3-clause BSD License.