Compute marginal effects at the mean or average marginal effects from statistical models and returns the result as tidy data frames. These data frames are ready to use with the 'ggplot2'-package. Marginal effects can be calculated for many different models. Interaction terms, splines and polynomial terms are also supported. The two main functions are ggpredict() and ggaverage(), however, there are some convenient wrapper-functions especially for polynomials or interactions. There is a generic plot()-method to plot the results using 'ggplot2'.
This package computes marginal effects at the mean or average marginal effects from statistical models and returns the result as tidy data frames. These data frames are ready to use with the ggplot2-package. Marginal effects can be calculated for many different models. Currently supported model-objects are:
gam (package mgcv),
svyglm.nb. Other models not listed here are passed to a generic predict-function and might work as well, or maybe with
ggeffect(), which effectively does the same as
Interaction terms, splines and polynomial terms are also supported. The two main functions are
ggaverage(), however, there are some convenient wrapper-functions especially for polynomials or interactions. There is a generic
plot()-method to plot the results using ggplot2.
The returned data frames always have the same, consistent structure and column names, so it's easy to create ggplot-plots without the need to re-write the function call.
predicted are the values for the x- and y-axis.
conf.high could be used as
ymax aesthetics for ribbons to add confidence bands to the plot.
group can be used as grouping-aesthetics, or for faceting.
ggpredict() requires at least one, but not more than three terms specified in the
terms-argument. Predicted values of the response, along the values of the first term are calucalted, optionally grouped by the other terms specified in
data(efc) fit <- lm(barthtot ~ c12hour + neg_c_7 + c161sex + c172code, data = efc) ggpredict(fit, terms = "c12hour") #> # A tibble: 62 × 6 #> x predicted conf.low conf.high group #> <dbl> <dbl> <dbl> <dbl> <fctr> #> 1 4 74.43040 72.33073 76.53006 1 #> 2 5 74.17710 72.09831 76.25588 1 #> 3 6 73.92379 71.86555 75.98204 1 #> 4 7 73.67049 71.63242 75.70857 1 #> 5 8 73.41719 71.39892 75.43546 1 #> 6 9 73.16389 71.16504 75.16275 1 #> 7 10 72.91059 70.93076 74.89042 1 #> 8 11 72.65729 70.69608 74.61850 1 #> 9 12 72.40399 70.46098 74.34700 1 #> 10 14 71.89738 69.98948 73.80529 1 #> # ... with 52 more rows
A possible call to ggplot could look like this:
library(ggplot2) mydf <- ggpredict(fit, terms = "c12hour") ggplot(mydf, aes(x, predicted)) + geom_line() + geom_ribbon(aes(ymin = conf.low, ymax = conf.high), alpha = .1)
However, there is also a
plot()-method. This method uses convenient defaults, to easily create the most suitable plot for the marginal effects.
mydf <- ggpredict(fit, terms = "c12hour") plot(mydf)
plot() offers a few, but useful arguments, so it's easy to use.
With three variables, predictions can be grouped and faceted.
ggpredict(fit, terms = c("c12hour", "c172code", "c161sex")) #> # A tibble: 372 × 7 #> x predicted conf.low conf.high group facet #> <dbl> <dbl> <dbl> <dbl> <fctr> <fctr> #> 1 4 74.70073 72.38031 77.02114 intermediate level of education  Female #> 2 4 73.98237 70.45711 77.50763 low level of education  Female #> 3 4 75.41908 71.91747 78.92070 high level of education  Female #> 4 4 73.65930 70.08827 77.23033 intermediate level of education  Male #> 5 4 72.94094 68.38540 77.49649 low level of education  Male #> 6 4 74.37766 70.05658 78.69874 high level of education  Male #> 7 5 74.44742 72.14644 76.74841 intermediate level of education  Female #> 8 5 73.72907 70.21926 77.23888 low level of education  Female #> 9 5 75.16578 71.67430 78.65726 high level of education  Female #> 10 5 73.40600 69.84575 76.96625 intermediate level of education  Male #> # ... with 362 more rows mydf <- ggpredict(fit, terms = c("c12hour", "c172code", "c161sex")) ggplot(mydf, aes(x = x, y = predicted, colour = group)) + stat_smooth(method = "lm", se = FALSE) + facet_wrap(~facet)
plot() works for this case, as well.
There are some more features, which are explained in more detail in the package-vignette.
The package is easily extendable, to add support for other model objects. The only requirement is that following methods are available:
family(). If model objects do not support these methods, you may implement workarounds (see below).
Following code needs to be revised to add further model objects:
get_model_function()needs a line to specify whether the new model can be considered as linear or generalized linear model.
get_predict_function()needs a line to specify the class.
select_prediction_method()to call the right prediction-method, and add a method
get_predictions_<class>(), if one of the existing prediction-methods does not fit the needs of the new model object.
When the model object does not support one of
family(), you may add workarounds:
family()-function, a workaround has to be added to
get_glm_family()in the file utils_model_family.R.
model.frame()-function with standard arguments or return values, a workaround has to be added to
get_model_frame()in the file utils_model_frame.R.
predict()-function, a workaround has to be added to
get_predictions_<class>()in the file predictions.R.
To install the latest development snapshot (see latest changes below), type following commands into the R console:
Please note the package dependencies when installing from GitHub. The GitHub version of this package may depend on latest GitHub versions of my other packages, so you may need to install those first, if you encounter any problems. Here's the order for installing packages from GitHub:
To install the latest stable release from CRAN, type following command into the R console:
In case you want / have to cite my package, please use
citation('ggeffects') for citation information.
convert_case()from sjlabelled, in preparation for the latest snakecase-package update.
brmsfit-models from the brms-package.
clm-models from the ordinal-package.
multinom-models from the nnet-package.
ppd) now compute uncertainty intervals also for non-gaussian models.
ggpredict()now computes the weighted mean as typical value for predictors that are held constant.
summary()function, to provide information on predictions by grouping variables, and on constant values from adjustments.
show.legend-argument to show or hide the legend of plots.
dot.alpha-argument, to specify a different alpha-values for data points when plotting raw data.
jitter-argument, to add a small amount of random variation to the location of data points when plotting raw data.
plot()and getter-functions (like
get_x_labels()) get a
case-argument, to convert labels into any case, using the snakecase-package.
betareg-models. Note, however, that due to some uncertainty, the intervals may not be "smooth".
ci.lvl) were not always recognized.
ggeffect(), if the term in question was categorical.
stanregmodels (pkg rstanarm).
plot()did not work for predictions at specific values (i.e. when certain levels of predictor where selected in square brackets).
mermod-objects did not work when model had only one fixed effects term.
polrmodels (pkg MASS).
zeroinflmodels (pkg pscl).
betaregmodels (pkg betareg).
truncregmodels (pkg truncreg).
coxphmodels (pkg survival).
emm()as convenient shortcut to compute the estimate marginal mean of the model's response value.
use.theme-argument, to use the default ggeffects-theme, or to use the default ggplot-theme.
ggpredict()computes proper confidence intervals for merMod- and lme-objects.
plot()-method, to better plot raw data.
rawdatadid not work for models with discrete binary response.