diff --git a/NEWS.md b/NEWS.md
index 6ce883c568..12c05d4dff 100644
--- a/NEWS.md
+++ b/NEWS.md
@@ -30,6 +30,8 @@
 
 3. In v1.12.4, fractional `fread(..., stringsAsFactors=)` was added. For example if `stringsAsFactors=0.2`, any character column with fewer than 20% unique strings would be cast as `factor`. This is now documented in `?fread` as well, [#4706](https://github.com/Rdatatable/data.table/issues/4706). Thanks to @markderry for the PR.
 
+4. `cube(DT, by="a")` now gives a more helpful error that `j` is missing, [#4282](https://github.com/Rdatatable/data.table/pull/4282).
+
 
 # data.table [v1.14.0](https://github.com/Rdatatable/data.table/milestone/23?closed=1)  (21 Feb 2021)
 
diff --git a/R/groupingsets.R b/R/groupingsets.R
index 6281615dd5..5c3ad02d4b 100644
--- a/R/groupingsets.R
+++ b/R/groupingsets.R
@@ -27,10 +27,12 @@ cube.data.table = function(x, j, by, .SDcols, id = FALSE, ...) {
     stop("Argument 'by' must be a character vector of column names used in grouping.")
   if (!is.logical(id))
     stop("Argument 'id' must be a logical scalar.")
+  if (missing(j))
+    stop("Argument 'j' is required")
   # generate grouping sets for cube - power set: http://stackoverflow.com/a/32187892/2490497
   n = length(by)
   keepBool = sapply(2L^(seq_len(n)-1L), function(k) rep(c(FALSE, TRUE), times=k, each=((2L^n)/(2L*k))))
-  sets = lapply((2L^n):1L, function(j) by[keepBool[j, ]])
+  sets = lapply((2L^n):1L, function(jj) by[keepBool[jj, ]])
   # redirect to workhorse function
   jj = substitute(j)
   groupingsets.data.table(x, by=by, sets=sets, .SDcols=.SDcols, id=id, jj=jj)
diff --git a/inst/tests/tests.Rraw b/inst/tests/tests.Rraw
index dfeb47841f..4535a6e048 100644
--- a/inst/tests/tests.Rraw
+++ b/inst/tests/tests.Rraw
@@ -17295,3 +17295,6 @@ test(2165.3, X[Y],                          data.table(A=2:3, B=2:3, i.A=2:1, ke
 test(2165.4, X[Y, on=.(A)],                 data.table(A=2:1, B=c(2L,NA), i.B=2:3))      # no key
 test(2165.5, X[Y, on=.(A), x.B, by=.EACHI], data.table(A=2:1, x.B=c(2L,NA)))             # no key
 
+# missing j was caught in groupingsets but not cube, leading to unexpected error message, #4282
+DT = data.table(a=1)
+test(2166, cube(DT, by='a'), error="Argument 'j' is required")