Dynamic Causal Model for Skin-Conductance Responses

Estimates trial-by-trial latent sudomotor input amplitudes that best explain the skin-conductance signal, in the spirit of the PsPM dynamic causal model for SCR (Bach et al., 2010, 2011). Each event contributes a non-negative latent input whose skin-conductance response is the canonical scrf kernel; the per-event amplitudes are estimated by non-negative least squares against a low-frequency cosine drift basis, and the input timings are optionally refined by a nonlinear local search (the "dynamic" step).

Usage

edaDCM(
  x,
  events,
  channel = 1,
  assay_name = NULL,
  tau1 = 0.75,
  tau2 = 2,
  drift = 6,
  refine_latency = TRUE,
  latency_window = 1.5
)

Arguments

x

A PhysioExperiment object with EDA data.

events

A data.frame with numeric onset (seconds); a condition column is optional and carried through.

channel

EDA channel index or label (default: 1).

assay_name

Input assay (default: defaultAssay(x)).

tau1, tau2

SCR rise/decay time constants in seconds.

drift

Number of cosine drift terms (default: 6).

refine_latency

Logical; refine each event's onset by a local search to improve the fit (default: TRUE).

latency_window

Half-width in seconds of the latency search (default: 1.5).

Value

A list of class "eda_dcm" with:

estimates

a data.frame with event, onset, onset_refined, amplitude (per-event latent input), and condition if present.

fitted

fitted signal; residuals; r_squared.

Details

Unlike edaGLM, which estimates one amplitude per condition, edaDCM estimates one latent input amplitude per event.

References

Bach, D.R., Daunizeau, J., Friston, K.J., & Dolan, R.J. (2010). "Dynamic causal modelling of anticipatory skin conductance responses." Biological Psychology, 85(1), 163-170. doi:10.1016/j.biopsycho.2010.06.007

See also

edaGLM for the condition-level GLM, scrf for the response kernel.

Examples

ev <- data.frame(onset = c(10, 30, 50, 70))
pe <- edaSimulate(n_time = 1000, sr = 10, scr_count = 4, seed = 1)
fit <- edaDCM(pe, ev, refine_latency = FALSE)
fit$estimates
#>   event onset onset_refined     amplitude
#> 1     1    10            10  2.864800e-01
#> 2     2    30            30 -2.775558e-17
#> 3     3    50            50 -3.112554e-17
#> 4     4    70            70  8.186079e-03