[R-sig-ME] glmer: object 'd' not found

[Steve Walker]

Martin Maechler is currently working on this issue:

https://github.com/lme4/lme4/issues/51#issuecomment-20554892

Steve

On 2013-07-06 12:58 AM, Ross Boylan wrote:
> Working with lme4 built from git today (92498e426de8086e4aef32cfc8c9995b6b2481d2):
>> r3b <- glmer(pAnyUAVI~(1|nid)+A5+rac_gay+nPartners+ethnic_sg+HIVTest, data=sp2a, family=binomial(), nAGQ=8)
> Error in stopifnot(length(d) == 2, length(sval) == p, diff(sval) <= 0) :
>    object 'd' not found
>
> Enter a frame number, or 0 to exit
>
> 1: glmer(pAnyUAVI ~ (1 | nid) + A5 + rac_gay + nPartners + ethnic_sg + HIVTest
> 2: eval(mc, parent.frame(1))
> 3: eval(expr, envir, enclos)
> 4: lme4::glFormula(formula = pAnyUAVI ~ (1 | nid) + A5 + rac_gay + nPartners +
> 5: rankMatrix(t(reTrms$Zt), method = "qr", sval = min(d))
> 6: stopifnot(length(d) == 2, length(sval) == p, diff(sval) <= 0)
>
> It looks as if there might be a problem in glFormula (modular.R), which includes this test
>      ## FIXME: adjust test for families with estimated scale parameter:
>      ##   useSc is not defined yet/not defined properly?
>      if (doCheck(cc <- control[[cstr]]) &&                   ## not NULL or "ignore"
>          !(grepl(cc,"Small") &&   ## not "*Small" and large Z mat
>            prod(d <- dim(reTrms$Zt)) > 1e6) &&
>          (nrow(fr) < (rankZ <- rankMatrix(t(reTrms$Zt), method="qr", sval=min(d))))) ## test
> I think the last line is the source of the call for level 5 of the stack.
> Theory: the value of the grepl test is false and so prod(d <- ...) is never evaluated, and d is undefined.
>
> I can't definitively confirm this in the debugger:
> Selection: 4
> Called from: stopifnot(length(d) == 2, length(sval) == p, diff(sval) <= 0)
> Browse[1]> class(reTrms$Zt)
> [1] "dgCMatrix"
> attr(,"package")
> [1] "Matrix"
> Browse[1]> dim(reTrms$Zt)
> [1] 1043 2505
> # not sure why there are so many columns
> Browse[1]> class(d)
> Error during wrapup: object 'd' not found
> #seems to indicate d undefined
> Browse[1]> prod(dim(reTrms$Zt))
> [1] 2612715
> # is > 1e6
> Browse[1]> cc
> [1] "stopSmall"
> Browse[1]> grepl(cc, "Small")
> [1] FALSE
>
> Selection: 5
> Called from: function ()
> {
>      if (.isMethodsDispatchOn()) {
>          tState <- tracingState(FALSE)
>          on.exit(tracingState(tState))
>    ...
> Browse[1]> sval
> Error during wrapup: promise already under evaluation: recursive default argument reference or earlier problems?
> # not sure what is going on there.  I did previously select 5 and got dumped back in the frame selection list
> # so maybe that messed things up.
>
> I think this exact same call worked with an earlier version of lme4 on Windows.
> Current environment R 2.15.1, Debian wheezy, amd64.  R and some packages are from the Debiain archive,
> many other dependencies of lme4 built from CRAN source, and lme4 from git.
>
> names(sp2a)
> [1] "pAnyUAVI"  "studyidx"  "A5"        "ethnic_sg" "HIVTest"   "rac_gay"
> [7] "nPartners" "1"         "nid"
> Note the "1" which is a column of 1's.  Perhaps that causes trouble with the (1|nid) specification?  But, again, I think this was
> working earlier.
>
> Ross Boylan
>
> _______________________________________________
> R-sig-mixed-models at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models