Skip to contents

Lagged PROTEST for several pairs of data sets

Source: R/pairs.R
lagged_protest_pairs.Rd

Runs lagged_protest() on pairs of data sets from a named list, such as the modules of a multi-trait study measured on the same time points, and collects the results in one table.

Usage

lagged_protest_pairs(
  data,
  pairs = NULL,
  lags = 0:6,
  bidirectional = TRUE,
  p.adjust.method = "BH",
  seed = NULL,
  ...
)

Arguments

data

A named list of numeric matrices, data frames or dist objects, all with the same number of rows (time points).

pairs

Pairs to analyze: NULL (default) for all unordered pairs, or a list of character vectors of length 2, or a two-column character matrix, giving names of data. The first element of each pair is passed as X.

lags, bidirectional

Passed to lagged_protest(). bidirectional defaults to TRUE here.

p.adjust.method

Correction of the global p-values across pairs, any method of stats::p.adjust(). Defaults to "BH".

seed

Passed to lagged_protest() for every pair.

...

Further arguments passed to lagged_protest(), such as permutation, nperm or period.

Value

A data frame with one row per pair and columns x, y, optimal_lag, n (rows at the optimal lag), m2, r, p.value (raw p at the optimal lag), p.global and p.global.adjusted (adjusted across pairs). The lagged_protest() fits are stored, named "x -> y", in the attribute "fits".

Details

By default every unordered pair is analyzed once with bidirectional = TRUE, so both directions are covered by the sign of the lag: a positive optimal lag means that the first data set of the pair (x) leads the second (y), and a negative one that y leads x. The global p-values of the pairs are then adjusted for the number of pairs with p.adjust.method, because the max-statistic test accounts for the lags scanned within a pair but not for the pairs examined.

Every pair uses the same seed, so results do not depend on the order or number of pairs.

See also

lagged_protest()

Examples

sim <- simulate_lagged_series(n_times = 12, true_lag = 2, noise_sd = 0.3,
                              seed = 1)
blocks <- list(A = sim$X, B = sim$Y,
               C = simulate_lagged_series(n_times = 12, seed = 2)$X)
res <- lagged_protest_pairs(blocks, lags = 0:3, nperm = 199, seed = 1)
res
#>   x y optimal_lag  n         m2         r p.value p.global p.global.adjusted
#> 1 A B           2 10 0.04963006 0.9748692   0.005    0.010            0.0300
#> 2 A C           3  9 0.21583858 0.8855289   0.095    0.275            0.2750
#> 3 B C           1 11 0.20289455 0.8928076   0.010    0.145            0.2175
plot(attr(res, "fits")[["A -> B"]])