Skip to contents

Two-wave change-score DMSA: time x methylation x factor

Source: R/change.R
dmsa_change.Rd

Tests Delta-S x exposure (x mod2) per unit through the composite machinery (dmsa_model, Freedman-Lane, family blocks), with maxT across the units of the level-local family on one shared permutation stream.

Usage

dmsa_change(
  M_before,
  M_after,
  data,
  outcome = NULL,
  exposure,
  mod2 = NULL,
  units,
  alignment,
  covariates = NULL,
  baseline = TRUE,
  role = c("predictor", "outcome"),
  block = NULL,
  B = 1999,
  correction = c("maxT", "minP"),
  weighting = c("combined", "reliability", "flat"),
  w_floor = 1.5,
  seed = 1
)

Arguments

M_before, M_after

numeric matrices, same probes in the same order (rows = the same people in the same order).

data

data.frame with outcome, exposure, moderators and covariates.

outcome

outcome column (role = "predictor").

exposure

the moderating factor E; with two waves this is the `mod` of the time x S x E three-way (Delta carries the time axis).

mod2

optional second moderator: tests Delta-S x E x mod2 (moderators mean-centered, all lower-order terms in the model, highest-order tested).

units

unit label per probe column (genes, modules, or one system).

alignment

alignment as in dmsa_triangulate.

covariates

character vector of covariate columns. Supply the CHANGE-SCALE cell composition here; a message reminds you if empty.

baseline

include the unit's before-wave score as a covariate.

role

"predictor": outcome ~ dS * E (* mod2) + S0 + covariates. "outcome": dS ~ E (* mod2) + covariates (exposure predicts the change).

block

exchangeable blocks (e.g. couple id).

B, correction, seed

as elsewhere; correction applies across the units

weighting

Character. Probe weighting engine within a unit: "combined" (default) fuses the flat and reliability statistics on one shared permutation stream, "flat" weights every usable aligned probe equally, "reliability" weights each probe by its item-rest correlation with the rest of its unit. Weights are computed from methylation alone, so the permutation null is unaffected.

w_floor

Numeric. Lower bound applied to reliability weights before normalisation, so a single poorly behaved probe cannot be driven to zero influence. Ignored when weighting = "flat". of the family (maxT default).

Value

data.frame of class "dmsa_change": unit, n, n_probes, b, t, p, p_adj; attributes term, role, formula.

Examples

set.seed(5)
n <- 80; g <- rep(c("g1", "g2"), each = 4); P <- length(g)
d <- rep(c(1, -1), length.out = P); E <- rnorm(n); y <- rnorm(n)
M0 <- matrix(rnorm(n * P), n, P)
M1 <- 0.6 * M0 + matrix(rnorm(n * P), n, P)
# plant a change in g1 that tracks y x E through the alignment
M1[, g == "g1"] <- M1[, g == "g1"] + outer(0.5 * y * E, d[g == "g1"])
al <- dmsa_align(data.frame(cpg = paste0("cg", 1:P), d = d,
                            p_plus = ifelse(d > 0, 0.9, 0.1)), genes = g)
dat <- data.frame(y = y, E = E, dEpi = rnorm(n))
dmsa_change(M0, M1, dat, outcome = "y", exposure = "E", units = g,
            alignment = al, covariates = "dEpi", B = 99, seed = 1)
#> DMSA change-score test - dS x E (time x S x E, 2-wave identity) 
#>   family of 2 unit(s), maxT across the family
#>   g1                           n   80  probes   4  b +0.507  t +6.15  p 0.0100  adj 0.0200  *
#>   g2                           n   80  probes   4  b -0.173  t -1.37  p 0.1600  adj 0.2900