dudi.hillsmith error in x * w : non-numeric argument to binary operator

Question

I'm re-running an RLQ & 4th corner analyses I coded about 6 months ago. However, an error is popping up with analysis of the Q table. I have been following Stephane Dray's (2013) tutorial and used

dudi.pca for the environment table (R, 6 "numeric" variables by 6 sites),

dudi.coa for the species presence table (L, 15 species ("integer" data) by 6 sites), and

dudi.hillsmith for the trait table (Q, 7 mixed traits (5 "character" and 2 "numeric") by 15 species).

The dudi.pca(R), dudi.coa(L) both run fine, but when I try:

> q_feeding <- dudi.hillsmith(traits_feeding, row.w=l_feeding$cw, scannf=F, nf=2)

I get

Error in x * w : non-numeric argument to binary operator

For simplicity of our discussion, "feeding" = "presence", it's a 0/1 as to whether the species was observed feeding at a site. This error also comes up with different data and when L is abundance data. I have tried changing the categorical traits to TRUE/FALSE, to see if "logical" would work. I'm stumped and hope it's something simple that I'm missing.

"row.w=l_feeding$cw" returns a vector of numbers, which tells me that the dudi.coa is working, so I am guessing my error lies within the "traits_feeding" table:

    > traits_feeding
          by    fe         dus      fnt    mo    se       ete
c.notat  0.450 0.673        Y        N     N      N        N
c.riat   0.423 0.667        Y        N     N      Y        N
z.op     0.501 0.556        N        Y     N      N        N
s.liat   0.335 0.596        N        Y     Y      N        N
s.i      0.382 0.549        N        Y     N      N        N
n.turat  0.354 0.648        N        Y     Y      N        N
c.eeke   0.297 0.348        Y        Y     N      Y        Y
c.rdid   0.306 0.688        Y        Y     N      Y        Y
s.tipinn 0.320 0.658        Y        Y     N      Y        N
s.amele  0.302 0.695        Y        Y     N      Y        N
s.midiat 0.285 0.678        Y        Y     N      Y        N
s.obb    0.294 0.666        Y        Y     N      Y        N
s.r      0.338 0.658        Y        Y     N      Y        N
s.vulat  0.315 0.684        Y        Y     N      Y        N
s.hlege  0.302 0.670        Y        Y     N      Y        N

> class(traits_feeding$by)
[1] "numeric"
> class(traits_feeding$fe)
[1] "numeric"
> class(traits_feeding$dus)
[1] "character"
> class(traits_feeding$fnt)
[1] "character"
> class(traits_feeding$mo)
[1] "character"
> class(traits_feeding$se)
[1] "character"
> class(traits_feeding$ete)
[1] "character"

I'm running R 4.0.0 on RStudio 1.2.5042 and am running the most recent (16-May-2020) of ade4.

Can anyone help me out with this? The most mystifying part is that this ran without error six months ago, I believe exactly as is here...

Thank you, Ryan

Answer 1

I faced the same problem as you and I believe that the problem is due to your R version and not related to the ade4 version. I was able to solve this problem by removing the R 4.0 and later installing the 3.6.2 version. Hope it helps.

Cheers!

Answer 2

I know that this is a rather late response, but here's what worked for me:

dudi.hillsmith cannot find a way to weigh character variables.

I changed these to factors with levels, and that worked perfectly fine.

Hope this helps anyone else with the same problem!