Tariff (version 1.0.5)

tariff: Replicate Tariff methods

Description

This function implements Tariff method.

Usage

tariff(causes.train, symps.train, symps.test, causes.table = NULL,
  use.rank = TRUE, nboot.rank = 1, use.sig = TRUE, nboot.sig = 500,
  use.top = FALSE, ntop = 40, ...)

Arguments

causes.train

character vector of causes, or the column name of cause in the training data

symps.train

N.train by S matrix

symps.test

N.test by S matrix

causes.table

list of causes in the data

use.rank

logical indicator for whether using ranks instead of scores

nboot.rank

number of re-sampling for baseline rank comparison. Default to 1, which resamples training data to have a uniform cause distribution of the same size. Set this to 0 removes bootstrapping the training dataset.

use.sig

logical indicator for whether using significant Tariff only

nboot.sig

number of re-sampling for testing significance.

use.top

logical indicator for whether the tariff matrix should be cleaned to have only top symptoms

ntop

number of top tariff kept for each cause

...

not used

Value

score

matrix of score for each cause within each death

causes.train

vector of most likely causes in training data

causes.test

vector of most likely causes in testing data

csmf

vector of CSMF

causes.table

cause list used for output, i.e., list of existing causes in the training data

use.rank

logical indicator for whether using ranks instead of scores

References

James, S. L., Flaxman, A. D., Murray, C. J., & Population Health Metrics Research Consortium. (2011). Performance of the Tariff Method: validation of a simple additive algorithm for analysis of verbal autopsies. Population Health Metrics, 9(1), 1-16.

Serina, P., Riley, I., Stewart, A., James, S. L., Flaxman, A. D., Lozano, R., ... & Ahuja, R. (2015). Improving performance of the Tariff Method for assigning causes of death to verbal autopsies. BMC medicine, 13(1), 1.

Tyler H. McCormick, Zehang R. Li, Clara Calvert, Amelia C. Crampin, Kathleen Kahn and Samuel J. Clark(2016) Probabilistic cause-of-death assignment using verbal autopsies, http://arxiv.org/abs/1411.3042 To appear, Journal of the American Statistical Association

Examples

# NOT RUN {
data("RandomVA3")
test <- RandomVA3[1:200, ]
train <- RandomVA3[201:400, ]
allcauses <- unique(train$cause)
fit <- tariff(causes.train = "cause", symps.train = train, 
				symps.test = test, causes.table = allcauses)
correct <- which(fit$causes.test[,2] == test$cause)
accuracy <- length(correct) / dim(test)[1]
# }