Fixing things up to pass devtools::check() in preparation for submiss…

…ion to CRAN

DESCRIPTION

@@ -2,17 +2,18 @@ Package: gFormulaMI
 Title: G-formula for causal inference via multiple imputation
 Version: 0.0.1
 Authors@R: 
-    person("Jonathan", "Bartlett", , "jonathan.bartlett1@lshtm.ac.uk", role = c("aut", "cre"),
-           comment = c(ORCID = "0000-0001-7117-0195"))
+    c(person("Jonathan", "Bartlett", , "jonathan.bartlett1@lshtm.ac.uk", role = c("aut", "cre"),
+           comment = c(ORCID = "0000-0001-7117-0195")),
+           person("Camila", "Olarte Parra", email = "camila.olarte-parra@lshtm.ac.uk", role = "ctb",
+           comment = c(ORCID = "0000-0003-0263-4392")))
 Description: gFormulaMI implements the G-formula method for causal inference
   using Bayesian multiple imputation methods. It creates multiple synthetic imputed
   datasets under treatment regimes of interest. These can then be analysed
   using rules developed for analysing multiple synthetic datasets.
 License: GPL (>= 3)
 Encoding: UTF-8
 Imports:
-  mice,
-  mitools
+  mice
 Roxygen: list(markdown = TRUE)
 RoxygenNote: 7.2.3
 Suggests: 

R/gFormulaImpute.R

@@ -37,6 +37,8 @@
 #' @param data The observed data frame
 #' @param M The number of imputed datasets to generate
 #' @param trtVars A vector of variable names indicating the time-varying treatment variables
+#' @param trtRegimes A vector specifying the treatment regime of interest, or a list of
+#' vectors specifying the treatment regimes of interest
 #' @param nSim The number of individuals to simulate in each imputed dataset. Defaults to
 #' number of individuals in observed data
 #' @param micePrintFlag TRUE/FALSE specifying whether the output from the call(s) to mice
@@ -78,7 +80,7 @@ gFormulaImpute <- function(data, M=50, trtVars, trtRegimes,
                            nSim=NULL, micePrintFlag=FALSE,silent=FALSE,
                            method=NULL,predictorMatrix=NULL) {
 
-  if (class(data)=="mids") {
+  if (inherits(data, "mids")) {
     missingData <- TRUE
     if (silent==FALSE) {
       print("Input data is a mice created multiple imputation object.")
@@ -91,7 +93,7 @@ gFormulaImpute <- function(data, M=50, trtVars, trtRegimes,
     }
     M <- data$m
     firstImp <- mice::complete(data,1)
-  } else if (class(data)=="data.frame")  {
+  } else if (inherits(data,"data.frame"))  {
     missingData <- FALSE
     if (silent==FALSE) {
       print("Input data is a regular data frame.")
@@ -208,6 +210,8 @@ gFormulaImpute <- function(data, M=50, trtVars, trtRegimes,
     }
 
     #remove original data from imputations
+    #but first declare regime=NULL to avoid CRAN note
+    regime <- NULL
     returnImps <- mice::filter(imps, regime!=0)
 
   } else {

R/syntheticPool.R

@@ -33,8 +33,8 @@
 #' @examples
 #' set.seed(7626)
 #' #impute synthetic datasets under two regimes of interest using gFormulaImpute
-#' imps <- gFormulaImpute(data=simDataFullyObs,M=10,trtVarStem="a",
-#'                         timePoints=3,
+#' imps <- gFormulaImpute(data=simDataFullyObs,M=10,
+#'                         trtVars=c("a0","a1","a2"),
 #'                         trtRegimes=list(c(0,0,0),c(1,1,1)))
 #' #fit linear model to final outcome with regime as covariate
 #' fits <- with(imps, lm(y~factor(regime)))
@@ -45,18 +45,18 @@
 syntheticPool <- function(fits) {
 
   M <- length(fits$analyses)
-  p <- length(fits$analyses[[1]]$coefficients)
+  p <- length(stats::coefficients(fits$analyses[[1]]))
   ests <- array(0, dim=c(M,p))
   within_vars <- array(0, dim=c(M,p))
 
   for (i in 1:M) {
-    ests[i,] <- coefficients(fits$analyses[[i]])
-    within_vars[i,] <- diag(vcov(fits$analyses[[i]]))
+    ests[i,] <- stats::coefficients(fits$analyses[[i]])
+    within_vars[i,] <- diag(stats::vcov(fits$analyses[[i]]))
   }
 
   overall_ests <- colMeans(ests)
   v <- colMeans(within_vars)
-  b <- diag(var(ests))
+  b <- diag(stats::var(ests))
   total_var <- (1+1/M)*b - v
 
   #check that all variances are positive
@@ -74,13 +74,13 @@ syntheticPool <- function(fits) {
   resTable[,4] <- total_var
   resTable[,5] <- df
   #95% confidence interval
-  resTable[,6] <- overall_ests - qt(0.975,df=df)*sqrt(total_var)
-  resTable[,7] <- overall_ests + qt(0.975,df=df)*sqrt(total_var)
+  resTable[,6] <- overall_ests - stats::qt(0.975,df=df)*sqrt(total_var)
+  resTable[,7] <- overall_ests + stats::qt(0.975,df=df)*sqrt(total_var)
   #two sided p-value
-  resTable[,8] <- 2*pt(-abs(overall_ests/sqrt(total_var)),df=df)
+  resTable[,8] <- 2*stats::pt(-abs(overall_ests/sqrt(total_var)),df=df)
 
   colnames(resTable) <- c("Estimate", "Within", "Between", "Total", "df",
                           "95% CI L", "95% CI U", "p")
-  rownames(resTable) <- names(coefficients(fits$analyses[[1]]))
+  rownames(resTable) <- names(stats::coefficients(fits$analyses[[1]]))
   resTable
 }