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

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

Description

Function for estimating the Potential Impact Fraction pif from a cross-sectional sample of the exposure X with a known Relative Risk function rr with parameters theta when only mean(X) and var(X) are known.

Usage

pif.approximate(X, Xvar, thetahat, rr, cft = NA, deriv.method.args = list(),
  deriv.method = c("Richardson", "complex"), check_exposure = TRUE,
  check_rr = TRUE, check_integrals = TRUE, is_paf = FALSE)

Arguments

X	Mean value of exposure levels from a cross-sectional random sample. If multivariate, this should be a matrix with 1 row and as many columns as covariates
Xvar	Variance of the exposure levels.
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	Twice differentiable function cft(X) for counterfactual. Leave empty for the Population Attributable Fraction paf where counterfactual is 0 exposure.
deriv.method.args	method.args for hessian.
deriv.method	method for hessian. Don't change this unless you know what you are doing.
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

pif.approximate method should be the last choice for the case when no information on the exposure X (except for mean and standard deviation) are given. In practice pif.empirical should be prefered.

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
#--------------------------------------------
X         <- data.frame(2)
thetahat  <- 0.12
Xvar      <- 0.2
rr        <- function(X,theta){exp(X*theta)}
cft       <- function(X){0.5*X}
pif.approximate(X, Xvar, thetahat, rr, cft)

#Change the derivative arguments 
pif.approximate(X, Xvar, thetahat, rr, cft, 
               deriv.method = "Richardson", 
               deriv.method.args = list(eps=1e-8, d=0.000001))

#When no counterfactual is specified paf is calculated
pif.approximate(X, Xvar, thetahat, rr)

#Example 2: Multivariate
#--------------------------------------------
X1        <- 2
X2        <- 1.1
X         <- data.frame(X1,X2)
Xvar      <- matrix(c(1,.4,.4,1),ncol = 2, byrow = TRUE)
cft       <- function(X){.25*X}
thetahat  <- c(0.12, 0.03)
rr        <- function(X, theta){exp(theta[1]*X[,1] + theta[2]*X[,2])}
pif.approximate(X, Xvar, thetahat, rr, cft)

#Example 3: More multivariate
#--------------------------------------------
X1       <- rnorm(1000,3,.5)
X2       <- rnorm(1000,4,1)
X        <- cbind(X1,X2)
Xmean    <- data.frame(t(colMeans(X)))
Xvar     <- var(X)
thetahat <- c(0.12, 0.17)
thetavar <- matrix(c(0.001, 0.00001, 0.00001, 0.004), byrow = TRUE, nrow = 2)
rr       <- function(X, theta){exp(theta[1]*X[,1] + theta[2]*X[,2])}
cft      <- function(X){cbind(sqrt(X[,1] + 0.2*X[,2]), X[,1])}
pif.approximate(Xmean, Xvar, thetahat, rr, cft)

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