pif.empirical: Point Estimate of the Potential Impact Fraction via the...

In pifpaf: Potential Impact Fraction and Population Attributable Fraction for Cross-Sectional Data

Description

Function that calculates the potential impact fraction pif via the empirical method. That is: for a random sample X, a relative risk function rr(X, thetahat) with parameters thetahat the empirical estimator is given by:

PIF = 1 - ∑ rr(cft(X_i); θ)/∑ rr(X_i; θ)

Usage

pif.empirical(X, thetahat, rr, cft = NA, weights = rep(1/nrow(as.matrix(X)),
  nrow(as.matrix(X))), check_exposure = TRUE, check_rr = TRUE,
  check_integrals = TRUE, is_paf = FALSE)

Arguments

X	Random sample (data.frame) which includes exposure and covariates. or sample mean if approximate method is selected.
thetahat	Estimator (vector) of theta for the Relative Risk function.
rr	Function for Relative Risk which uses parameter theta. The order of the parameters shound be rr(X, theta). **Optional**
cft	Function cft(X) for counterfactual. Leave empty for the Population Attributable Fraction paf where counterfactual is 0 exposure.
weights	Normalized survey weights for the sample X.
check_exposure	Check that exposure X is positive and numeric.
check_rr	Check that Relative Risk function rr equals 1 when evaluated at 0.
check_integrals	Check that counterfactual and relative risk's expected values are well defined for this scenario.
is_paf	Boolean forcing evaluation of paf.

Note

The empirical method converges for relative risk rr functions that are Lipschitz, convex or concave on thetahat. For stranger functions use pif.kernel.

Author(s)

Rodrigo Zepeda-Tello [e-mail]

Dalia Camacho-Garc<c3><ad>a-Forment<c3><ad> [e-mail]

See Also

pif which is a wrapper for all pif methods (pif.empirical, pif.approximate, pif.kernel).

For estimation of the Population Attributable Fraction see paf.

Examples

#Example 1: Relative risk given by exponential
#--------------------------------------------
set.seed(18427)
X        <- data.frame(rnorm(100,3,.5))
thetahat <- 0.12
rr       <- function(X, theta){exp(theta*X)}
pif.empirical(X, thetahat, rr, cft = function(X){ 0.5*X })

#Without counterfactual estimates PAF
pif.empirical(X, thetahat, rr) 
 
#Example 2: Linear relative risk
#--------------------------------------------
pif.empirical(X, thetahat, rr = function(X, theta){theta*X + 1}, 
               cft = function(X){ 0.5*X })

#Example 3: Multivariate relative risk
#--------------------------------------------
set.seed(18427)
X1       <- rnorm(100,4,1)
X2       <- rnorm(100,2,0.4)
X        <- data.frame(cbind(X1,X2))
thetahat <- c(0.12, 0.03)
rr       <- function(X, theta){exp(theta[1]*X[,1] + theta[2]*X[,2])}

#Creating a counterfactual. As rr requires a bivariate input, cft should 
#return a two-column matrix
cft  <- function(X){
   cbind(X[,1]/2, 1.1*X[,2])
}
pif.empirical(X, thetahat, rr, cft) 

pifpaf documentation built on May 1, 2019, 9:11 p.m.