CLR: na.rm is not working

[TuomasBorman]

vegan/R/decostand.R

Line 149 in 69de595

.calc_clr <-

Hi,

I noticed that na.rm is not used in CLR calculation.

-Tuomas

[jarioksa]

There is a bigger problem than this. See my comments with #663. All these related standardizations share this same problem: users cannot set na.rm, but na.rm = TRUE will be always used (though with "clr" only after your suggested fix).

Here a dissection of the problem:

1. decostand and vegdist have argument na.rm that defaults FALSE.
2. method clr delegates calculations to .calc_clr, but it does not pass the value of na.rm set in decostand call. (It passes dot arguments, but na.rm is not among them.)
3. .calc_clr sets na.rm = TRUE independent of settings in decostand or vegdist.

[jarioksa]

I am not sure about the internal logic of these functions. I need comment by @antagomir

Currently these methods work with hard-coded na.rm = TRUE in decostand. Looking at the code, it may be that this should be the case. Comments in .calc_rclr say so. If this is case, the calls to .calc_* functions should drop na.rm = TRUE from the function call and explicitly write na.rm = TRUE where needed. Only if there are cases where NA could optionally be retained, should we use na.rm = na.rm, with na.rm defined at the function call. Further, to make na.rm user settable, the .calc_* functions should be called as, say, .calc_clr(x, na.rm = na.rm, ...) where na.rm is set at the decostand call.

I started to make changes in these functions but got scared of fooling up, and I'll leave this till I hear from @antagomir .

[antagomir]

It would be better to activate na.rm for users. I can update this but will also check if we could finalize the matrix completion step to the rclr method. Will come back this week.

[antagomir]

@TuomasBorman kindly check #667 ?