Take the consensus of two Minimum Data Set selection routes

Runs both the PCA and the correlation-network selection over the same data and returns the indicators both agree on. This is a cheap and underused robustness check: an indicator selected by two methods that reward different properties – variance and centrality – is more likely to matter than one selected by either alone.

Usage

mds_consensus(data, pca_args = list(), na_args = list(), standardize = TRUE)

Arguments

data

A data frame of indicator values.

pca_args

A named list of arguments passed to pca_select_mds. Note that the PCA route expects standardised data.

na_args

A named list of arguments passed to na_select_mds.

standardize

If TRUE (the default), the data are standardised with standardize_numeric before the PCA route, which requires it. The network route is given the unstandardised data, since Spearman correlation is invariant to monotonic rescaling and standardising would make no difference.

Value

A list with:

consensus

Character vector of indicators selected by both routes

pca

The full pca_select_mds result

network

The full na_select_mds result

pca_only

Indicators selected only by PCA

network_only

Indicators selected only by the network route

Details

On Yuan and Shi's (2026) 40-year tillage data, soil organic carbon, dissolved organic carbon and soil compaction were selected by all six minimum-data-set variants tested.

The intersection can legitimately be empty. That is informative, not an error: it means the two routes disagree completely about which indicators carry the signal, which is a reason to look at the data before trusting either.

See also

pca_select_mds, na_select_mds

Examples

if (requireNamespace("igraph", quietly = TRUE)) {
  props <- c("Sand", "Silt", "Clay", "pH", "OM", "SOC", "N", "P", "K",
             "CEC", "Ca", "Mg", "BD", "EC")
  agreement <- mds_consensus(soil_data[, props])

  agreement$consensus
  agreement$pca_only
  agreement$network_only
}
#> character(0)