paf.confidence.inverse: Confidence intervals for the Population Attributable... In pifpaf: Potential Impact Fraction and Population Attributable Fraction for Cross-Sectional Data

Description

Confidence intervals for the Population Attributable Fraction for relative risk inyective functions, the PAF is inyective, and intervals can be calculated for the relative risk, and then transformed to PAF CI.

Usage

 ```1 2 3 4 5``` ```paf.confidence.inverse(X, thetahat, rr, thetavar, weights = rep(1/nrow(as.matrix(X)), nrow(as.matrix(X))), method = c("empirical", "approximate"), nsim = 1000, confidence = 95, deriv.method.args = list(), deriv.method = c("Richardson", "complex"), force.min = FALSE, check_thetas = TRUE, Xvar = var(X)) ```

Arguments

 `X` Random sample (`data.frame`) which includes exposure and covariates. `thetahat` Estimative 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** `thetavar` Estimator of variance of `thetahat`. `weights` Survey `weights` for the random sample `X`. `method` Either `empirical` (default) or `approximate`. `nsim` Number of simulations (default: `1000`) `confidence` Confidence level % (default: `95`) `deriv.method.args` `method.args` for `hessian`. Only if `"approximate"` method is chosen. `deriv.method` `method` for `hessian`. Don't change this unless you know what you are doing. Only if `"approximate"` method is chosen. `force.min` Boolean indicating whether to force the `rr` to have a minimum value of 1 instead of 0 (not recommended). `check_thetas` Checks that theta parameters are correctly inputed `Xvar` Variance of exposure levels.

Note

The `force.min` option forces the relative risk `rr` to have a minimum of `1` and thus an `rr < 1` is NOT possible. This is only for when absolute certainty is had that `rr > 1` and should be used under careful consideration. The confidence interval to acheive such an `rr` is based on the paper by Do Le Minh and Y. .s. Sherif

Author(s)

Rodrigo Zepeda-Tello [email protected]

Dalia Camacho-Garc<c3><ad>a-Forment<c3><ad> [email protected]

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 38``` ``` ## Not run: #Example 1: Exponential Relative Risk #-------------------------------------------- set.seed(18427) X <- as.data.frame(rnorm(100,0.3,.05)) thetahat <- 0.4 thetavar <- 0.1 paf.confidence.inverse(X, thetahat, function(X, theta){exp(theta*X)}, thetavar) #With approximate method Xmean <- as.data.frame(mean(X[,1])) Xvar <- var(X) paf.confidence.inverse(Xmean, thetahat, function(X, theta){exp(theta*X)}, thetavar, Xvar = Xvar, method = "approximate") #We can force PAF's CI to be >= 0 (only if it is certain) paf.confidence.inverse(X, thetahat, function(X, theta){exp(theta*X)}, thetavar, force.min = TRUE) #Example 2: Multivariate Relative Risk #-------------------------------------------- set.seed(18427) X1 <- rnorm(1000,0.3,.05) X2 <- rnorm(1000,0.3,.05) X <- as.data.frame(as.matrix(cbind(X1,X2))) thetahat <- c(0.12, 0.03) thetavar <- matrix(c(0.1, 0, 0, 0.4), byrow = TRUE, nrow = 2) rr <- function(X, theta){exp(theta[1]*X[,1] + theta[2]*X[,2])} paf.confidence.inverse(X, thetahat, rr, thetavar) #Same example with approximate method Xmean <- as.data.frame(matrix(colMeans(X), ncol = 2)) Xvar <- cov(X) paf.confidence.inverse(Xmean, thetahat, rr=rr, thetavar = thetavar, method = "approximate", Xvar = Xvar) ## End(Not run) ```

pifpaf documentation built on Sept. 29, 2017, 1:03 a.m.