provides different diagnostic plots for an aggregation procedure.
# S3 method for mixture
plot(
x,
pause = FALSE,
col = NULL,
alpha = 0.01,
dynamic = T,
type = "all",
max_experts = 50,
col_by_weight = TRUE,
...
)an object of class mixture. If awake is provided (i.e., some experts are unactive), their residuals and cumulative losses are computed by using the predictions of the mixture.
if set to TRUE (default) displays the plots separately, otherwise on a single page
the color to use to represent each experts, if set to NULL (default) use RRColorBrewer::brewer.pal(...,"Spectral"
numeric. Smoothing parameter for contribution plot (parameter 'f' of function lowess).
boolean. If TRUE, graphs are generated with rAmCharts, else with base R.
char.
'all' Display all the graphs ;
'plot_weight', 'boxplot_weight', 'dyn_avg_loss', 'cumul_res', 'avg_loss', 'contrib' Display the selected graph alone.
integer. Maximum number of experts to be displayed (only the more influencial).
boolean. If TRUE (default), colors are ordered by weights of each expert, else by column
additional plotting parameters
plots representing: plot of weights of each expert in function of time, boxplots of these weights, cumulative loss \(L_T=\sum_{t=1}^T l_{i,t}\) of each expert in function of time, cumulative residuals \(\sum_{t=1}^T (y_t-f_{i,t})\) of each expert's forecast in function of time, average loss suffered by the experts and the contribution of each expert to the aggregation \(p_{i,t}f_{i,t}\) in function of time.
See opera-package and opera-vignette for a brief example about how to use the package.