pif.kernel: Kernel-based estimate of Potential Impact Fraction

Description Usage Arguments Value Note Author(s) See Also Examples

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 using kernels.

Usage

1
2
3
4
5
6
pif.kernel(X, thetahat, rr, cft = NA, weights = rep(1/nrow(as.matrix(X)),
  nrow(as.matrix(X))), adjust = 1, n = 512, ktype = c("gaussian",
  "epanechnikov", "rectangular", "triangular", "biweight", "cosine",
  "optcosine"), bw = c("SJ", "nrd0", "nrd", "ucv", "bcv"),
  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.

adjust

Adjust bandwith parameter from density from density.

n

Number of equally spaced points at which the density is to be estimated (see density).

ktype

kernel type: "gaussian", "epanechnikov", "rectangular", "triangular", "biweight", "cosine", "optcosine". Additional information on kernels in density

bw

Smoothing bandwith parameter from density from density. Default "SJ"

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.

Value

pif Estimate of Potential Impact Fraction

Note

In practice pif.empirical should be prefered as convergence is faster than pif.kernel for most functions. In addition, the scope of pif.kernel is limited as it does not work with multivariate exposure data X.

Author(s)

Rodrigo Zepeda-Tello rzepeda17@gmail.com

Dalia Camacho-GarcĂ­a-FormentĂ­ daliaf172@gmail.com

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.

For more information on kernels see density

Examples

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
#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.kernel(X, thetahat, rr, cft = function(X){ 0.5*X })

#Choose a different kernel
pif.kernel(X, thetahat, rr, cft = function(X){ 0.5*X }, ktype = "gaussian")

#Specify kernel options
pif.kernel(X, thetahat, rr, cft = function(X){ 0.5*X }, ktype = "gaussian", 
bw = "nrd", adjust = 0.5, n = 1100)

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

#Example 3: More complex counterfactual
#--------------------------------------------
set.seed(18427)
X       <- data.frame(rnorm(100,4,1))
thetahat <- c(0.12, 0.03)
rr       <- function(X, theta){1 + theta[1]*X + theta[2]*X^2}

#Creating a counterfactual. As rr requires a bivariate input, cft should 
#return a two-column matrix
cft  <- function(X){
   X[which(X > 4)] <- 1
   return(X)
}
pif.kernel(X, thetahat, rr, cft) 

INSP-RH/pif documentation built on May 7, 2019, 6:01 a.m.