evalmodels {RainTyrol} | R Documentation |
The function evalmodels
fits distributional trees (disttree
),
distributional forests (distforest
),
a prespecified GAMLSS (gamlss
),
a boosted GAMLSS (gamboostLSS
),
and an EMOS model (crch
) to precipitation data.
The results are compared based on CRPS, log-likelihood and RMSE.
evalmodels(station, train, test, ntree = 100, distfamily = "gaussian", tree_minsplit = 50, tree_minbucket = 20, tree_mincrit = 0.95, forest_minsplit = 50, forest_minbucket = 20, forest_mincrit = 0, forest_mtry = 27, gamboost_cvr = FALSE)
station |
|
train |
|
test |
|
ntree |
|
distfamily |
|
tree_minsplit |
|
tree_mincrit |
|
tree_minbucket |
|
forest_minsplit |
|
forest_minbucket |
|
forest_mincrit |
|
forest_mtry |
|
gamboost_cvr |
|
evalmodels
returns a list with the following components:
CRPS |
CRPS (continuos ranked probability score) of all methods, average over testing data. |
LS |
Logarithmic score (= log-likelihood) of all methods, average over testing data. |
RMSE |
Root mean squared error of all methods, average over testing data. |
if(require("crch") & require("disttree") & require("gamlss") & require("gamlss.dist") & require("gamlss.cens") & require("gamboostLSS") & require("mboost") & require("partykit") & require("scoringRules") & require("survival") ) { evalmodels(station = "Axams", train = 1985:2008, test = 2009:2012, distfamily = "gaussian") }