paf.sensitivity: Population Attributable Fraction Sensitivity Analysis Plot

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

Description

Function that plots a sensitivity analysis for the Population Attributable Fraction paf by checking how estimates vary when reducing the exposure's sample X.

Usage

paf.sensitivity(X, thetahat, rr, method = "empirical",
  weights = rep(1/nrow(as.matrix(X)), nrow(as.matrix(X))), nsim = 50,
  mremove = min(nrow(as.matrix(X))/2, 100), adjust = 1, n = 512,
  ktype = "gaussian", bw = "SJ", ylab = "PAF",
  xlab = "Number of randomly deleted observations for X",
  legendtitle = "Sensitivity Analysis",
  title = paste0("Population Attributable Fraction (PAF) ",
  "Sensitivity Analysis"), colors = c("red", "deepskyblue", "gray75",
  "gray25"), check_exposure = TRUE, check_rr = TRUE,
  check_integrals = TRUE)

Arguments

X	Random sample (data.frame) which includes exposure and covariates or sample mean if "approximate" method is selected.
thetahat	Asymptotically consistent or Fisher consistent estimator (vector) of theta for the Relative Risk function.
rr	function for Relative Risk which uses parameter theta. The order of the parameters should be rr(X, theta). **Optional**
method	Either "empirical" (default), "kernel" or "approximate". For details on estimation methods see pif.
weights	Normalized survey weights for the sample X.
nsim	Integer with number of samples to include (for each removal) in order to conduct sensitivity analysis. See details for additional information.
mremove	Limit number of measurements of X to remove when resampling. See details for additional information.
adjust	Adjust bandwith parameter (for "kernel" method) from density.
n	Number of equally spaced points at which the density (for "kernel" method) is to be estimated (see density).
ktype	kernel type: "gaussian", "epanechnikov", "rectangular", "triangular", "biweight", "cosine", "optcosine" (for "kernel" method). Additional information on kernels in density.
bw	Smoothing bandwith parameter (for "kernel" method) from density. Default "SJ".
ylab	string label for the Y-axis of the plot.
xlab	string label for the X-axis of the plot.
legendtitle	String title for the legend of plot.
title	string Title of plot.
colors	String vector with colors for the plot.
check_exposure	boolean Check that exposure X is positive and numeric.
check_rr	boolean Check that Relative Risk function rr equals 1 when evaluated at 0.
check_integrals	boolean Check that counterfactual cft and relative risk's rr expected values are well defined for this scenario.

Details

paf.sensitivity conducts a sensitivity analysis of the paf estimate by removing mremove elements nsim times and re-estimating paf with the reduced sample.

Value

plotpaf ggplot object plotting a sensitivity analysis of paf.

Author(s)

Rodrigo Zepeda-Tello [e-mail]

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

See Also

paf for Population Attributable Fraction estimation, paf.plot for a plot of Population Attributable Fraction as a function of the relative risk's parameter theta.

Examples

## Not run: 
#Example 1
#------------------------------------------------------------------
set.seed(3284)
X  <- data.frame(rnorm(250,3))            #Sample
rr <- function(X,theta){exp(X*theta)}     #Relative risk
theta <- 0.1                              #Estimate of theta
paf.sensitivity(X, thetahat = theta, rr = rr)


#Save file
#require(ggplot2)
#ggsave("My Population Attributable Fraction Sensitivity Analysis.pdf")

#Example 2
#--------------------------------------------------------------
set.seed(3284)
X     <- data.frame(rbeta(1000, 1, 0.2))
theta <- c(0.12, 1)
rr    <- function(X, theta){X*theta[1] + theta[2]}


#Using empirical method
paf.sensitivity(X, thetahat = theta, rr = rr, 
                mremove = 100, nsim = 50, 
                title = "My Sensitivity Analysis for example 1")
                
#Same example with kernel
paf.sensitivity(X, theta, rr = rr, 
                 mremove = 100, nsim = 50, method = "kernel", 
                 title = "Sensitivity Analysis for example 1 using kernel")
                 

#Example 3: Plot counterfactual with categorical risks
#------------------------------------------------------------------
set.seed(18427)
X        <- data.frame(Exposure = 
               sample(c("Normal","Overweight","Obese"), 1000, 
                      replace = TRUE, prob = c(0.4, 0.1, 0.5)))
thetahat <- c(1, 1.7, 2)

#Categorical relative risk function
rr <- function(X, theta){

   #Create return vector with default risk of 1
   r_risk <- rep(1, nrow(X))
   
   #Assign categorical relative risk
   r_risk[which(X[,"Exposure"] == "Normal")]      <- thetahat[1]
   r_risk[which(X[,"Exposure"] == "Overweight")]  <- thetahat[2]
   r_risk[which(X[,"Exposure"] == "Obese")]       <- thetahat[3]
   
   return(r_risk)
}


pafplot <- paf.sensitivity(X, thetahat = thetahat, rr = rr, 
                           title = "Sensitivity analysis of PAF for excess-weight",
                           colors = rainbow(4), 
                           legendtitle = "Values", 
                           check_exposure = FALSE, check_rr = FALSE)              
pafplot              

#You can edit pafplot as it is a ggplot object
#require(ggplot2)
#pafplot + theme_classic()

## End(Not run)

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