diagnosability: Diagnosability

In EricArcher/strataG: Summaries and Population Structure Analyses of Genetic Data

Diagnosability

Description

Conduct Random Forest on a gtypes object to compute the diagnosability of each stratum (PD from Archer et al 2017).

Usage

diagnosability(
  g,
  gene = 1,
  pairwise = FALSE,
  conf.level = 0.95,
  replace = FALSE,
  sampsize = NULL,
  train.pct = 0.5,
  min.n = 2,
  min.votes.pct = c(0.8, 0.9, 0.95),
  rp.nrep = 0,
  unk = NULL
)

Arguments

g	haploid gtypes object with aligned sequences.
gene	number or name of gene to use from multidna @sequences slot. Defaults to the first gene in the object.
pairwise	do analysis on all pairwise combinations of strata?
conf.level	confidence level for the binom.test confidence interval.
replace	sample with replacement in Random Forest trees? (see randomForest).
sampsize	sample size for each Random Forest tree? (see randomForest). If NULL a balanced sample size is chosen (see balancedSampsize).
train.pct	if sampsize is NULL, the percent of the minimum strata size to use for sampsize.
min.n	minimum sample size across all strata.
min.votes.pct	numeric vector giving the minimum percent of votes for the assigned strata for a sample to be considered correctly assigned.
rp.nrep	number of replicates for rfPermute computation of significance of site importance scores.
unk	vector of strata to be treated as "unknowns" for prediction with Random Forest model.

Value

a list containing a data.frame of summary statistics (smry), and the randomForest object (rf). If pairwise is TRUE then the rf element is a list of randomForest results for each row in smry.

Author(s)

Eric Archer [e-mail]

Examples

## Not run: 
library(strataG)
data(dloop.g)

pd <- diagnosability(dloop.g, pairwise = TRUE)

lapply(pd, function(x) x$rf.confusion.mat)

## End(Not run)

EricArcher/strataG documentation built on June 8, 2025, 2:12 a.m.