# GEEmediate: Mediation Analysis for Generalized Linear Models Using the... In GEEmediate: Mediation Analysis for Generalized Linear Models Using the Difference Method

## Description

Estimation of natural direct and indirect effects for generalized linear models. The function utilizes a data-duplication algorithm to fit marginal and conditional GLMs in a way that allow for consistent variance estimation. The function produces point estimates, confidence intervals and p-values for the natural indirect effect and the mediation proportion

## Usage

 ```1 2 3``` ```GEEmediate(formula, exposure, mediator, df, family = gaussian, corstr = "independence", conf.level = 0.95, surv = F, pres = "sep", niealternative = "two-sided", ...) ```

## Arguments

 `formula` A formula expression as for other regression models, of the form response ~ predictors. See the documentation of `lm` and `formula` for details. predictors should include exposure/treatment and mediator. `exposure` The exposure (string). `mediator` The mediator (string). `df` A name of a data frame where all variables mentioned in formula are stored. `family` A `family` object to be used in gee: a list of functions and expressions for defining link and variance functions see the gee documentation. Default is gaussian. See also `gee` and `glm`. `corstr` A working correlation structure. See `gee` and `glm`. `conf.level` Confidence level for all confidence intervals (default 0.95) `surv` Is the outcome survival (not supported) `pres` Presentation of the coefficient tables. "tog" for a single table, "sep" for two separated tables. `niealternative` Alternative hypothesis for testing that the nie=0. Either "two-sided" (default) or "one-sided" for alternative nie>0. `...` Further arguments for the `gee` call.

## Value

The output contains the following components:

 `call` The call. `GEE.fit` Results of fitting the GEE for the duplicated data. `nie` The natural indirect effect estimate. NIE and NDE are reported on the coefficient scale `nie.pval` P-value for tesing mediation using the NIE. `nde` The natural direct effect estimate. `nie.ci` Confidence interval in for the NIE in confidence level conf.level. `pm` The mediation proportion estimate. `pm.pval` P-value for tesing one-sided mediation using the mediation proportion. `pm.ci` Confidence interval for the mediation proportion in confidence level conf.level.

## References

Nevo, Liao and Spiegelman, Estimation and infernece for the mediation proportion (2017+)

## Examples

 ``` 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18``` ```## Not run: SimNormalData <- function(n,beta1.star = 1, p = 0.3, rho =0.4, inter = 0) { beta2 <- (p/rho)*beta1.star beta1 <- (1-p)*beta1.star XM <- MASS::mvrnorm(n, mu = c(0,0), Sigma = matrix(c(1,rho,rho,1),2,2)) X <- XM[,1] M <- XM[,2] beta <- c(inter, beta1, beta2) print(beta) Y <- cbind(rep(1,n),XM)%*%beta+rnorm(n,0,sd = 1) return(data.frame(X = X, M = M, Y = Y)) } set.seed(314) df <- SimNormalData(500) GEEmediate(Y ~ X + M, exposure = "X", mediator = "M", df = df) ## End(Not run) ```

GEEmediate documentation built on July 18, 2019, 9:08 a.m.