Skip to contents

Parameter Sensitivity Analysis

Source: R/sensitivity_analysis.R
sensitivity_analysis.Rd

Reruns codiv() across a grid of span_fraction and min_symbiont_tips values and summarizes how the number of scanned and significant nodes changes, so you can see whether findings are stable or threshold-dependent.

Usage

sensitivity_analysis(
  Host_tree,
  Symbiont_tree,
  Host_to_Symbiont_df,
  span_fraction_range = c(0.05, 0.1, 0.2, 0.3),
  min_symbiont_tips_range = c(5, 7, 10, 15),
  permutations = 99,
  min_hosts = 3,
  methods = c("hommola", "paco", "parafit"),
  alpha = 0.05,
  cores = 1,
  seed = 8675309,
  verbose = TRUE
)

Arguments

Host_tree

A phylo object; binary phylogenetic tree of hosts

Symbiont_tree

A phylo object; binary phylogenetic tree of symbionts

Host_to_Symbiont_df

A data frame with columns "Host" and "Symbiont"

span_fraction_range

Numeric vector of span_fraction values to test; default c(0.05, 0.1, 0.2, 0.3)

min_symbiont_tips_range

Integer vector of min_symbiont_tips values; default c(5, 7, 10, 15)

permutations

Integer; number of permutations, applied to every run

min_hosts

Integer; minimum hosts, applied to every run (>= 3)

methods

Character vector of methods to use

alpha

Significance threshold for counting significant nodes; default 0.05

cores

Number of cores passed to codiv()

seed

Random seed for reproducibility

verbose

If TRUE (default), report progress across the grid

Value

A list with elements:

  • sensitivity_data: one row per parameter combination, with the number of nodes scanned and, per method, the number of significant nodes and the mean of the primary statistic

  • results_by_params: named list of the full codiv() result for each parameter combination

  • heatmap_data: long-format data (one row per combination and method) with n_sig, n_nodes, and prop_sig, ready for plotting

Details

Runtime is roughly baseline_time x length(span_fraction_range) x length(min_symbiont_tips_range). Tree-distance metrics are turned off for speed, since sensitivity is judged from node counts and significance.

Examples

# \donttest{
sim <- simulate_codiv_data(n_hosts = 10, n_clades = 3, seed = 1)
sens <- sensitivity_analysis(sim$host_tree, sim$symbiont_tree, sim$links,
                             span_fraction_range = c(0.3, 0.6),
                             min_symbiont_tips_range = c(7, 10),
                             permutations = 99, methods = "hommola",
                             verbose = FALSE)
sens$sensitivity_data
#>   span_fraction min_symbiont_tips n_nodes hommola_n_sig hommola_mean_stat
#> 1           0.3                 7      10             0       -0.05226943
#> 2           0.6                 7      12             0       -0.03900951
#> 3           0.3                10       8             0        0.01857405
#> 4           0.6                10      10             0        0.02031726
# }