pif.heatmap: Graphical Sensitivity Analysis of Potential Impact Fraction's...

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

Description

Provides a graphical sensitivity analysis for pif by varying the parameters of a bivariate counterfactual function cft. By default it evaluates the counterfactual:

cft(X) = aX + b.

Usage

pif.heatmap(X, thetahat, rr, cft = function(X, a, b) {     a * X + b },
  method = "empirical", weights = rep(1/nrow(as.matrix(X)),
  nrow(as.matrix(X))), Xvar = var(X), deriv.method.args = list(),
  deriv.method = "Richardson", adjust = 1, n = 512, ktype = "gaussian",
  bw = "SJ", legendtitle = "PIF", mina = 0, maxa = 1, minb = -1,
  maxb = 0, nmesh = 10,
  title = paste0("Potential Impact Fraction (PIF) with counterfactual",
  "\nf(X)= aX+b"), xlab = "a", ylab = "b",
  colors = rev(heat.colors(nmesh)), 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 shound be rr(X, theta). **Optional**
cft	function cft(X, a, b) for counterfactual dependent on one dimensional parameters a and b. Default counterfactual is affine: aX + b.
method	Either "empirical" (default), "kernel" or "approximate". For details on estimation methods see pif.
weights	Normalized survey weights for the sample X.
Xvar	Variance of exposure levels (for "approximate" method).
deriv.method.args	method.args for hessian (for "approximate" method).
deriv.method	method for hessian. Don't change this unless you know what you are doing (for "approximate" method).
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".
legendtitle	string title for the legend of plot.
mina	Minimum for parameter a for the counterfactual (default 0).
maxa	Maximum for parameter a for the counterfactual (default 1).
minb	Minimum for parameter b for the counterfactual (default -1).
maxb	Maximum for parameter b for the counterfactual (default 0).
nmesh	Number of tiles in plot (default 10).
title	string title for the plot.
xlab	string label for the X-axis of the plot (corresponding to "a").
ylab	string label for the Y-axis of the plot (corresponding to "b").
colors	vector of colours for the heatmap.
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.

Value

plotpif ggplot object plotting a heatmap with sensitivity analysis of the counterfactual.

Author(s)

Rodrigo Zepeda-Tello [e-mail]

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

See Also

See pif for Potential Impact Fraction estimation, pif.sensitivity for sensitivity analysis of the convergence process, pif.plot for a plot of Potential Impact Fraction as a function of the relative risk's parameter theta.

Examples

## Not run: 
#Example 1
#------------------------------------------------------------------
X  <- data.frame(rnorm(25,3))            #Sample
rr <- function(X,theta){exp(X*theta)}    #Relative risk
theta <- 0.01                            #Estimate of theta
pif.heatmap(X, theta = theta, rr = rr)

#Save file using ggplot2
#require(ggplot2)
#ggsave("My Potential Impact Fraction Heatmap Analysis.pdf")

#Change pif estimation method to kernel

pif.heatmap(X, theta = theta, rr = rr, method = "kernel")

#Example 2
#------------------------------------------------------------------
X     <- data.frame(Exposure = rbeta(100, 1, 0.2))
theta <- c(0.12, 1)
rr    <- function(X,theta){X*theta[1] + theta[2]}
cft   <- function(X, a, b){sin(a*X)*b}
pif.heatmap(X, theta = theta, rr = rr, cft = cft, 
     nmesh = 15, colors = rainbow(30), method = "empirical",
     title = "PIF with counterfactual cft(X) = sin(a*X)*b")

#Change estimation method to approximate
Xmean <- data.frame(mean(X[,"Exposure"]))
Xvar  <- var(X)
pif.heatmap(Xmean, Xvar = Xvar, theta = theta, rr = rr, cft = cft, 
     nmesh = 15, colors = rainbow(30), method = "approximate",
     title = "PIF with counterfactual cft(X) = sin(a*X)*b")


#Example 3: Plot univariate counterfactuals
#------------------------------------------------------------------
X       <- data.frame(rgamma(100, 1, 0.2))
theta   <- c(0.12, 1)
rr      <- function(X,theta){X*theta[1] + theta[2]}
cft     <- function(X, a, b){sqrt(a*X)}   #Leave two variables in it
pifplot <- pif.heatmap(X, theta = theta, rr = rr, 
 cft = cft, mina = 0, maxa = 1, minb = 0, maxb = 0, 
 legendtitle = "Potential Impact Fraction",
 title ="Univariate counterfactual", ylab = "", colors = topo.colors(10))
pifplot

#You can also add additional ggplot objects
#require(ggplot2)
#pifplot + annotate("text", x = 0.25, y = 0.4, label = "10yr scenario") + 
#geom_vline(aes(xintercept = 0.5), linetype = "dashed") +
#geom_segment(aes(x = 0.25, y = 0.38, xend = 0.5, yend = 0.3), 
#arrow = arrow(length = unit(0.25, "cm")))

#Example 4: Plot counterfactual with categorical risks
#------------------------------------------------------------------
set.seed(18427)
X        <- data.frame(Exposure = 
              sample(c("Normal","Overweight","Obese"), 100, 
                      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)
}

#Counterfactual of reducing a certain percent of obesity and overweight cases
#to normality
cft <- function(X, per.over, per.obese){

   #Find the overweight and obese individuals
   which_obese <- which(X[,"Exposure"] == "Obese")
   which_over  <- which(X[,"Exposure"] == "Overweight")
   
   #Reduce per.over % of overweight and per.obese % of obese
   X[sample(which_obese, length(which_obese)*per.obese),
       "Exposure"] <- "Normal"
   X[sample(which_over,  length(which_over)*per.over),
       "Exposure"] <- "Normal"
   
   return(X)
}

pif.heatmap(X, thetahat = thetahat, rr = rr, cft = cft, mina = 0, minb = 0, maxa = 1, 
            maxb = 1, title = "PIF of excess-weight reduction",
            xlab = "% Overweight cases", ylab = "% Obese cases",
            check_exposure = FALSE, check_rr = FALSE)

## End(Not run)

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