added_da: Test added diagnostic accuracy (estimated with AUC) dued to...

In lbraglia/lbdiag: Diagnostic Accuracy Miscellaneous Utilities

Test added diagnostic accuracy (estimated with AUC) dued to variable insertion in a logistic model

Description

The function add one terms per iteration, estimate the model, calculates the roc and finally compare them with z test using delong variance (by default, but other tests can be asked for, using pROC::roc.test).

Usage

added_da(
  formula,
  data,
  perm_test_sequence = FALSE,
  plot_roc = TRUE,
  col = "black",
  lty = 1,
  legend_style = c("covariates_auc", "numbers_auc", "none"),
  test_style = c("all_combs", "BvsC", "AvsC"),
  test_param = list()
)

Arguments

formula	a glm full formula
data	data.frame
plot_roc	plot the overlapping rocs of the models estimated
col	color of the rocs
lty	lty of the rocs
legend_style	covariates_auc for the right hand formula and auc, numbers_auc for number between parenthesis and auc, none for avoiding printing legend
test_style	all_combs for testing all the combinations, AvsC for comparing all the models to the one with only one covariate, BcsC to compare the models with the one identical excepting for the last covariate added
test_param	list of parameter to be passed to pROC::roc.test
perm_tests_sequence	permutate model covariate sequence (checking all possible models/sequences given the covariates specified)

Source

idea in Knottnerus Buntinx (editors), Evidence base of clinical diagnosis, 2008, Wiley Blackwell, pag 157

lbraglia/lbdiag documentation built on July 29, 2023, 3:35 p.m.