A proposed feature is to add a Normal option to eda_qq whereby eda_qq generates a Normal QQ plot. In such a case, x is a vector y is NULL and a parameter norm is set to TRUE.

The show.par parameter is ignored when generating a qq plot using the eda_qq function. But, it's respected when setting md=TRUE.

eda_qq will assign the names x and y to the interpolated batches in the output list. Instead, it would be best to assign the input variable names.

It seems that eda_rline is ignoring the iter argument in its iteration. For example,

eda_rline(neoplasms, Temp, Mortality, iter =1)

will still iterate until a stable slope is achieved.

...
$px
[1] 1
$py
[1] 1
$iter
[1] 4
attr(,"class")
[1] "eda_rline"

Split batch into two halves using median.

med <- median(x)
len <- length(x)
x <- sort(x)
n2 <- ifelse( len%%2 == 0, len/2, (len + 1)/2)
vi <- med - x[1:n2]
ui <- x[ (len + 1) - (1:n2) ] - med

eda_qq(vi,ui)

ref: Sections 2.3 and 2.8 of Graphical methods for data analysis (Chambers et al.)

eda_rline's tukey argument is set to TRUE by default. It should be set to FALSE so as to match the default settings in other functions.

Add axes labels to eda_boxls.
Streamline its "looks" to match those of the other plots.

eda_boxls has an unused argument, outliers. This is not to be confused with the outlier argument which is used in the function.

The dplyr function in eda_sl raises a Note in the package check (this fails the check in TravisCI). A possible solution follows:

  df2 <- df1[order(df1$grp, df1$y),]
  df3 <- split(df1, df1$grp)
  y <- lapply(df3, function(x) as.vector(x[,2]) )
  n <- lapply(df3, FUN = function(x)nrow(x) )
  M <- lapply(n, function(x) floor((x - 1) / 2))
  H <- lapply(M, function(x) (floor(x) - 1 ) / 2)
  med <- lapply(1:length(y), FUN=function(x, lst1, lst2) log(lst1[[x]][lst2[[x]]]) ,
                lst1=y, lst2=M)
  Hlo <- lapply(1:length(y), FUN=function(x, lst1, lst2) lst1[[x]][floor(lst2[[x]])] ,
                lst1=y, lst2=H)
  n_hi <- lapply(1:length(M), function(x, lst1, lst2) ceiling(lst1[[x]] + 1 - lst2[[x]]),
                 lst1 = n, lst2 = H)
  Hhi <- lapply(1:length(y), FUN=function(x, lst1, lst2) lst1[[x]][lst2[[x]] + 1] ,
                lst1=y, lst2=n_hi)

  df4 <- data.frame( grp = unique(df2$grp), med = unlist(med), Hhi = unlist(Hhi),
                    Hlo = unlist(Hlo))
  df4$sprd <- log(df4$Hhi - df4$Hlo)
  return(df4)

The title should be more explicit and emphasize that the function generates an empirical QQ plot (as opposed to a theoretical QQ plot such as a Normal QQ plot).

The Tukey transformation needs to take on the form of -x^p when p < 0.

Many functions use side=3 in the mtext function to add the y-axis label. This needs to be changed to side=2, for example:

mtext(ylab, side = 2,  col = plotcol, padj = -0.5, at = par('usr')[4], las = 2)

This appears to be related to this issue. As of dplyr 0.8, if dplyr is not attached, n() must be called via dplyr::n() inside the summary() function.

Per https://r-pkgs.org/r.html, make sure that calls to par() are properly reset via something like:

old <- options(stringsAsFactors = FALSE)
on.exit(options(old), add = TRUE)

The Tukey re-expression equation needs to be modified in the help page.