ErrorCheck: ErrorCheck

In SharonLutz/Umediation: Runs the Umediation Function

Description

Performs error checks on the input for the Umediation function.

Usage

   ErrorCheck<-function(n=1000,Atype="D",Mtype="C",Ytype="C",Ctype="C",Utype="C",interact=FALSE,muC=0,varC=1,muU=0,varU=1,
                        gamma0=0,gammaC=0,gammaU=0,varA=1,alpha0=0,alphaA=0,alphaC=0,alphaU=0,varM=1,beta0=0,betaA=0,betaM=0,
                        betaI=0,betaC=0,betaU=0,varY=1,alpha=0.05,nSim=300,nBoot=500)

Arguments

n	is the sample size of the population that is being simulated.
Atype	is either "C" for continuous, normally distributed exposure A or "D" for dichotomous, binary exposure A (i.e. Atype=”D”).
Mtype	is either "C" continuous, normally distributed mediator M or "D" for dichotomous, binary mediator M (i.e. Mtype=”C”).
Ytype	is either "C" for continuous, normally distributed outcome Y or "D" for dichotomous, binary outcome Y (i.e. Ytype=”D”).
Ctype	is either "C" for continuous, normally distributed measured confounder C or "D" for dichotomous, binary measured confounder C. Ctype can be a single value (i.e. Ctype=”C”) or a vector for multiple measured confounders (i.e. Ctype=c(”C”,”C”,”D”) ).
Utype	is either "C" for continuous, normally distributed unmeasured confounder U or "D" for dichotomous, binary unmeasured confounder U. Utype can be a single value (i.e. Utype=”C”) or a vector for multiple measured confounders (i.e. Utype=c(”C”,”C”,”D”) ).
interact	Using the flag interact=TRUE allows for an interaction between the exposure A and the mediator M on the outcome Y (i.e. E[Y]=beta0+betaA*A+betaM*M+betaC*C+betaU*U+betaI*A*M). By default, interact=FALSE (i.e. E[Y]=beta0+betaA*A+betaM*M+betaC*C+betaU*U).
muC	is the mean vector for the measured confounder C. For continuous measured confounder (i.e. Ctype=”C”), muC is the mean of C. For dichotomous measured confounder C (i.e. Ctype=”D”), muC is the probability C=1.
varC	is the variance of the measured confounder C when Ctype=”C”. For multiple measured confounders, the length of varC must match muC and Ctype (i.e. Ctype=c(”C”,”C”,”D”) and muC=c(-0.1,0.2,0.3) and varC=c(1,1,1))
muU	is the mean vector for the unmeasured confounder U. For continuous unmeasured confounder (i.e. Utype=”C”), muU is the mean of U. For dichotomous unmeasured confounder U (i.e. Utype=”D”), muU is the probability U=1.
varU	is the variance of the unmeasured confounder U when Utype=”C”. For multiple unmeasured confounders, the length of varU must match muU and Utype (i.e. Utype=c(”C”,”C”,”D”) and muU=c(-0.1,0.2,0.3) and varU=c(1,1,1))
gamma0	specifies the intercept for the exposure A (i.e. logit(P(A=1)) or E[A]=gamma0+gammaU*U+gammaC*C).
gammaC	specifies the relationship between the measured confounder C and the exposure A (i.e. logit(P(A=1)) or E[A]=gamma0+gammaU*U+gammaC*C).
gammaU	specifies the relationship between the measured confounder U and the exposure A (i.e. logit(P(A=1)) or E[A]=gamma0+gammaU*U+gammaC*C).
varA	is the variance of the exposure A when Atype=”C”. Default is varA=1.
alpha0	specifies the intercept for the mediator M (i.e. Logit(P(M=1)) or E[M]=alpha0+alphaA*A+alphaC*C+alphaU*U).
alphaA	specifies the relationship between the exposure A and the mediator M (i.e logit(P(M=1)) or E[M]=alpha0+alphaA*A+alphaC*C+alphaU*U).
alphaC	specifies the relationship between the measured confounder C and the mediator M (i.e. E[M] or logit(P(M=1))=alpha0+alphaA*A+alphaC*C+alphaU*U).
alphaU	specifies the relationship between the unmeasured confounder U and the mediator M (i.e. E[M] or logit(P(M=1))=alpha0+alphaA*A+alphaC*C+alphaU*U).
varM	is the variance of the mediator M when Mtype=”C”. Default is varM=1.
beta0	specifies the intercept for the outcome Y (i.e. logit(P(Y=1)) or E[Y]=beta0+betaA*A+betaM*M+betaI*A*M+betaC*C+betaU*U).
betaA	specifies the relationship between the exposure A and the outcome Y (i.e. E[Y] or logit(P(Y=1))=beta0+betaA*A+betaM*M+betaI*A*M+betaC*C+betaU*U).
betaM	specifies the relationship between the mediator M and the outcome Y (i.e. E[Y] or logit(P(Y=1))=beta0+betaA*A+betaM*M+betaI*A*M+betaC*C+betaU*U).
betaI	specifies the interaction between the mediator M and the exposure A on the outcome Y (i.e. E[Y] or logit(P(Y=1))= beta0+betaA*A+betaM*M+betaI*A*M+betaC*C+betaU*U) when the flag interact=TRUE.
betaC	specifies the relationship between the measured confounder C and the outcome Y (i.e. E[Y] or logit(P(Y=1))=beta0+betaA*A+betaM*M+betaI*A*M+betaC*C+betaU*U).
betaU	specifies the relationship between the unmeasured confounder U and the outcome Y (i.e. E[Y] or logit(P(Y=1))=beta0+betaA*A+betaM*M+betaI*A*M+betaC*C+betaU*U).
varY	is the variance of the outcome Y when Ytype=”C”. Default is varY=1.
alpha	is the significance level. Default value is alpha=0.05.
nSim	is the number of simulations run for the function. The more simulations run, the more accurate the results, but this will make the function slower.
nBoot	is the number of Monte Carlo draws for nonparametric bootstrap or quasi-Bayesian approximation for the mediate function.
seed	sets the seed used for the random generator.
seed	sets the seed used for the random generator.
atreat	sets the treatment group for the exposure A.
acontrol	sets the control group for the exposure A.

Value

The function exits with an error message if the error checks are not met.

Author(s)

Sharon Lutz, Annie Thwing

SharonLutz/Umediation documentation built on May 28, 2021, 6:39 a.m.