ord_plot fails after CLR if there are only two features/taxa

[nr0cinu]

Hi,

here is a minimal example of the issue:

ibd |> 
  tax_select(c("Firmicutes", "Bacteroidetes", "Proteobacteria")) |>
  tax_transform("clr", rank = "Phylum") |>
  ord_calc() |>  
  ord_plot() 
# works

ibd |> 
  tax_select(c("Firmicutes", "Bacteroidetes")) |>
  tax_transform("clr", rank = "Phylum") |>
  ord_calc() |>  
  ord_plot() 
# Error in ordsum[["sites"]][, axes, drop = FALSE] : 
#  subscript out of bounds

Thanks!

[david-barnett]

Hi,

Thanks for the clear, easy to repro example!

In this example, ord_calc runs a PCA, via vegan rda, that returns only one dimension.
This must be because clr makes the two taxa perfectly collinear.

For now, If you only need to compare two taxa, you can skip CLR and use a simple log transform after zero replacement instead, that keeps both variables independent and the PCA should then show two axes as expected.

I’ll add a clearer message or warning about this case in a future version to make it less confusing/surprising.

cheers,
David

library(microViz)
#> microViz version 0.12.7 - Copyright (C) 2021-2025 David Barnett
#> ! Website: https://david-barnett.github.io/microViz
#> ✔ Useful?  For citation details, run: `citation("microViz")`
#> ✖ Silence? `suppressPackageStartupMessages(library(microViz))`

# collinear output of clr transformation
ibd |> 
  tax_select(c("Firmicutes", "Bacteroidetes")) |>
  tax_transform("clr", rank = "Phylum") |>
  otu_get(samples = 1:5)
#> OTU Table:          [2 taxa and 5 samples]
#>                      taxa are columns
#>        Firmicutes Bacteroidetes
#> 099A  0.810619815  -0.810619815
#> 199A -0.099984743   0.099984743
#> 062B -0.017464502   0.017464502
#> 194A -0.006268441   0.006268441
#> 166A  0.166613412  -0.166613412

# single PCA axis output
ibd |> 
  tax_select(c("Firmicutes", "Bacteroidetes")) |>
  tax_transform("clr", rank = "Phylum") |>
  ord_calc() |> 
  ord_get()
#> Call: rda(formula = OTU ~ 1, data = data)
#> 
#> -- Model Summary --
#> 
#>               Inertia Rank
#> Total           3.353     
#> Unconstrained   3.353    1
#> 
#> Inertia is variance
#> 
#> -- Eigenvalues --
#> 
#> Eigenvalues for unconstrained axes:
#>   PC1 
#> 3.353

^{Created on 2025-10-06 with reprex v2.1.1}

[nr0cinu]

Thanks for explaining! Then we will add extra logic to our workflows to deal with this appropriately :)