WIER_SHARED_NNLS_model_setup: Model template based on Wierda et al.(2012)'s model

In JoKra1/papss: What the package does (short line)

Model template based on Wierda et al.(2012)'s model

Description

Creates trainings matrix and penalties for a fully penalized version of the original Wierda et al. (2012) model. Like by Wierda et al. separate sets of coefficients are estimated per factor level with the h_basis terms. All those coefficients are again constrained to be non-negative. The Model also includes the slope terms for each factor level that Wierda et al. (2012) introduced. These are not constrained - since their primary purpose in the experiments by Wierda et al. was to account for drift (i.e., positive or negative) in the pupil averages of individual subjects.

This model differs from the one by Wierda et al. (2012) in that it penalizes the coefficients corresponding to the h_basis terms. Specifically, it enforces a single penalty term shared by all factor levels (e.g., this is similar to the 'fs' basis in mgcv, Wood, 2017). The form of the penalty expressed on all of the basis functions is a simple identity matrix. We here also penalize all slope terms, again with a single penalty.

Usage

WIER_SHARED_NNLS_model_setup(
  expanded_time,
  expand_by,
  time,
  fact,
  pulse_locations,
  n,
  t_max,
  f
)

Arguments

expanded_time	A numeric vector containing positive time values in ms, expanded by a certain amount of ms
expand_by	Expansion time in ms passed to papss::pupil_solve(expand_by=) divided by sample length in ms
time	A numeric vector containing positive time values in ms
fact	The factor column from the data-frame passed to papss::pupil_solve()
pulse_locations	A numeric vector containing index values of pulse loc.
n	Parameter defined by Hoeks & Levelt (number of laters)
t_max	Parameter defined by Hoeks & Levelt (response maximum in ms)
f	Parameter defined by Wierda et al. (scaling factor)

Details

See: Wierda, S. M., van Rijn, H., Taatgen, N. A., & Martens, S. (2012). Pupil dilation deconvolution reveals the dynamics of attention at high temporal resolution. Proceedings of the National Academy of Sciences of the United States of America, 109(22), 8456–8460.

For penalty setup see: Wood, S. N. (2017). Generalized Additive Models: An Introduction with R, Second Edition (2nd ed.). Chapman and Hall/CRC.

JoKra1/papss documentation built on June 15, 2022, 8:57 a.m.