make_gaugepreds: Make gauge predictions

In messamat/globalIRmap: What the Package Does (One Line, Title Case)

Usage

make_gaugepreds(
  in_rftuned,
  in_gaugestats,
  in_res_spcv = NULL,
  in_predvars,
  interthresh,
  simple = FALSE
)

Arguments

in_rftuned	trained random forest model, output from selecttrain_rf
in_gaugestats	data.table of formatted gauging station summary statistics and hydro-environmental attributes. Output from format_gaugestats.
in_res_spcv	ResampleResult from spatial cross-validation, output from dynamic_resamplebm.
in_predvars	data.table of predictor variable codes, names and attributes. See selectformat_predvars.
interthresh	(numeric) between 0 and 1 (inclusive), probability threshold above which to classify gauging stations as non-perennial (e.g., 0.50).
simple	(logical) if TRUE, return only final model predictions, if FALSE return cross-validation predictions

Value

in_gaugestats data.table with one new column (IRpredprob_full, if simple == T) or five new columns (if simple == F).

• IRpredprob_full: predicted probability that station is non-perennial based on final model training.

• IRpredprob_CVsp: predicted probability that station is non-perennial based on spatial cross-validation (i.e., prediction when fold whose station is a member was excluded from training)

• IRpredprob_CVnosp: predicted probability that station is non-perennial based on spatial cross-validation

• inter_class_u10_featsel_spfold orinter_class_o1_featsel_spfold: cross-validation fold membership (id) }

Make predictions for all gauging stations used in model training based on final trained model. Join outputs of prediction and cross-validations to original gauging stations statistics table.

messamat/globalIRmap documentation built on July 4, 2021, 10:48 a.m.