Using svycontrast inside a function when contrasts involve backticks and I()

I am trying to write an R function implementing the approach to estimating kurtosis in the survey package, following the discussion here:

svykurt <- function(
    x,
    design,
    na.rm = FALSE,
    excess = TRUE
) {
  
  if (!inherits(design, "survey.design"))
    stop("design is not a survey design")
  
  # Storing variable in formula as string
  # This is necessary to construct the  moments formula below
  var_name <- as.character(x)[2]
  
  x <- model.frame(x, design$variables, na.action = na.pass)
  x <- as.matrix(x)
  
  if(na.rm){
    x <- x[!is.na(x)]
  }
  
  momnts_fmla <- paste0(
    "~",
    var_name,
    " + I(",
    var_name,
    "^2) + I(",
    var_name,
    "^3) + I(",
    var_name,
    "^4)")
  
  momnts <- svymean(
    as.formula(momnts_fmla),
    design,
    na.rm = na.rm
  )
  
  centrl_momnts <- svycontrast(
    momnts,
    list(
      mu4 = bquote(-3 * .(x^4) + 6 * .(x^2) * .(`I(x^2)`) - 4 * .(x) * .(`I(x^3)`) + .(`I(x^4)`)),
      sigma2 = bquote(.(`I(x^2)`) - .(x^2))
    )
  )
  
  return(centrl_momnts)
  
}

I am stuck on using svycontrast to transform raw moment into central moments. I found a helpful discussion here, but still run into trouble when passing things like I(x^2) to bquote inside the contrasts argument of svycontrast.

data(api)
dclus1 <- svydesign(id = ~dnum, weights = ~pw, data = apiclus1, fpc = ~fpc)
svykurt(x = ~api00, dclus1, na.rm = TRUE)

Error in eval(e[[2L]], where) : object 'I(x^2)' not found

If anyone has an idea on how I could approach this, let me know.

Thank you!

I think it's easier to rename the variable at the beginning. For example

svykurt<-function(formula, design){
	var_name <- formula[[2]]
	if (".x." %in% colnames(design)) stop("🙄")
	design<-eval(bquote(update(design, .x.=.(var_name))))
	moments<-svymean(~.x.+I(.x.^2)+I(.x.^3)+I(.x.^4), design)
	central_moments<-svycontrast(moments, 
		list(mu4=quote(-3*.x.^4+6*.x.^2*`I(.x.^2)`-4*.x.*`I(.x.^3)`+`I(.x.^4)`),
			 sigma2=quote(`I(.x.^2)`-`.x.`^2)
	))
	svycontrast(central_moments, quote(mu4/(sigma2*sigma2)))	
}

This is obviously not perfect, because someone might have a variable called .x..

An alternative, which breaks the opaqueness of classes, is to rename the elements of moments

svykurt1<-function(formula, design){
	var_name <- as.character(formula[[2]])
	moments_formula <- as.formula(paste0(
    "~",
    var_name,
    " + I(",
    var_name,
    "^2) + I(",
    var_name,
    "^3) + I(",
    var_name,
    "^4)"))
  
	moments<-svymean(moments_formula, design)
	attr(moments,"names")<-c("one","two","three","four")
	dimnames(attr(moments,"var"))<-list(c("one","two","three","four"),c("one","two","three","four"))
	central_moments<-svycontrast(moments, 
		list(mu4=quote(-3*one^4+6*one^2*two-4*one*three+four),
			 sigma2=quote(two-one^2)
	))
	svycontrast(central_moments, quote(mu4/(sigma2*sigma2)))	
}

Giving

> svykurt1(~api00, dclus2)
          nlcon     SE
contrast 1.7765 0.2284
> svykurt(~api00, dclus2)
          nlcon     SE
contrast 1.7765 0.2284

This doesn't answer the question of how you quote a name that contains backticks, which does indeed seem hard. You can work around it another way by deferring the translation from character to name. Here we construct the expressions that go into svycontrast by substituting character strings into an expression. The character strings don't need extra quoting because they are already dead.

This is probably the most elegant of the three solutions:

svykurt2<-function(formula, design){
	var_name <- as.character(formula[[2]])
	moments_char<-paste0("I(",var_name,"^",1:4,")")
	moments_formula <- make.formula(moments_char)
	moments<-svymean(moments_formula, design)
	mu4expr<-substitute(-3*one^4+6*one^2*two-4*one*three+four, 
			list(one=as.name(moments_char[1]),two=as.name(moments_char[2]),
			three=as.name(moments_char[3]),four=as.name(moments_char[4])))
	sigma2expr<-substitute(two-one^2, 
			list(one=as.name(moments_char[1]),two=as.name(moments_char[2])))
	central_moments<-svycontrast(moments, 
		list(mu4=mu4expr,
			 sigma2=sigma2expr
	))
	svycontrast(central_moments, quote(mu4/(sigma2*sigma2)))	
}

and it gives the right answer

> svykurt2(~api00, dclus2)
          nlcon     SE
contrast 1.7765 0.2284

PS: Wht hv y gt gnst vwls?