PsPM-style GLM for Skin-Conductance Responses

Fits a general linear model of the skin-conductance signal against event-related regressors, each a stick function at the event onsets convolved with the canonical skin-conductance response function (scrf; Bach & Friston, 2013). Events are grouped into conditions; each condition yields an amplitude estimate (beta) with a standard error and t statistic. A low-frequency cosine drift basis absorbs the tonic level. Optionally, a temporal-derivative basis is added per condition to accommodate response latency differences.

Usage

edaGLM(
  x,
  events,
  conditions = NULL,
  channel = 1,
  assay_name = NULL,
  tau1 = 0.75,
  tau2 = 2,
  basis = c("canonical", "derivative"),
  drift = 6
)

Arguments

x

A PhysioExperiment object with EDA data.

events

A data.frame with numeric onset (seconds) and a condition column grouping events.

conditions

Optional character vector selecting/ordering the conditions (default: the unique conditions in events).

channel

EDA channel index or label to model (default: 1).

assay_name

Input assay (default: defaultAssay(x)).

tau1, tau2

SCR rise/decay time constants in seconds.

basis

"canonical" (SCRF only) or "derivative" (SCRF plus its temporal derivative per condition).

drift

Number of cosine drift terms (default: 6).

Value

A list of class "eda_glm" with:

coefficients

a data.frame with condition, beta, se, t, p (canonical amplitude per condition; NA for a condition dropped due to collinearity).

derivative

(basis = "derivative") the derivative-regressor betas.

fitted

fitted signal; residuals; sigma; df; r_squared; rank; n_regressors; collinear.

References

Bach, D.R., & Friston, K.J. (2013). "Model-based analysis of skin conductance responses." Psychophysiology, 50(1), 15-22. doi:10.1111/j.1469-8986.2012.01483.x

See also

edaDCM for trial-wise latent input estimation, scrf for the response kernel.

Examples

ev <- data.frame(onset = c(10, 30, 50, 70), condition = c("A", "A", "B", "B"))
pe <- edaSimulate(n_time = 1000, sr = 10, scr_count = 4, seed = 1)
fit <- edaGLM(pe, ev)
fit$coefficients
#>   condition        beta         se         t            p
#> 1         A  0.04408197 0.02909868  1.514913 1.301132e-01
#> 2         B -0.13266133 0.02937474 -4.516170 7.053846e-06