Graphical heuristic for choosing the drift parameter in tfa()

choose_lambda runs temporal factor analyses for drift parameters in a specified range, and ouputs the percentage of variance of sample time explained by all factors for each value.

choose_lambda(model, Y, min_range = -5, max_range = 1,
  grid_size = 10, plot_res = TRUE, detail_axis = FALSE)

Arguments

model

an object of class 'tfa' adjusted with tfa().
Y

an nxp numeric matrix containing genetic information for n individuals recorded in p columns. Genetic information could be encoded by any numeric value, not necessarily an integer value. Missing data are not allowed.
min_range

a numeric value for the minimal range of drift parameter tested. Log 10 scale. The default value is min_range = -5.
max_range

a numeric value for the maximal range of drift parameter tested. Log 10 scale. The default value is max_range = 1.
grid_size

an integer value for the number of drift parameter tested. The default is grid_size = 10.
plot_res

a logical indicating whether the results should be displayed graphically or not. The default value is TRUE.
detail_axis

a logical indicating whether the results should also be displayed for each factor independently. The default value is FALSE.

Value

A vector of percentages of variance of sample time explained by factors for each value of the drift parameter in the specified range.

Details

This function requires that a preliminary model has been adjusted with K factors, where K is the number of source populations. The curve exhibits a drop for some value in the range, and the heuristic consists of choosing the largest values corresponding to removing time variation in factor K.

References

François, O., Liégeois, S., Demaille, B., Jay, F. (2019). Inference of population genetic structure from temporal samples of DNA. bioRxiv, 801324. https://www.biorxiv.org/content/10.1101/801324v3

François, O., Jay, F. (2020). Factor analysis of ancient population genomic samples. Under review.

See also

England_BA, tfa

Examples

library(tfa)

# Ancient DNA from Bronze Age Great Britain samples
data(england_ba)
attach(England_BA)

coverage <- meta$Coverage
geno <- coverage_adjust(genotype, coverage, K = 4, log = TRUE)

# Adjust an FA model
mod  <- tfa(age,
            geno,
            k = 3,
            lambda = 5e-1,
            center = TRUE,
            coverage = coverage,
            log = TRUE)

r_2 <- choose_lambda(mod,
                     geno,
                     min_range = -4,
                     max_range = 3)
abline(v=log10(5e-1), col = "orange", lty = 2)

# Remove HG from Serbia to keep k = 2 ancestral populations

age2 <- age[meta$Group.ID != "Serbia_HG"]
geno2 <- geno[meta$Group.ID != "Serbia_HG",]

# Adjust an FA model
mod  <- tfa(age2,
            geno2,
            k = 2,
            lambda = 5e-2,
            center = TRUE)

r_2 <- choose_lambda(mod,
                     geno2,
                     min_range = -3,
                     max_range = 2)
abline(v=log10(5e-2), col = "orange", lty = 2)
detach(England_BA)
# rm(list = ls())