m_step

m_step(N, Z, signewsq, xnew, a, muone, startflag, binflag)

M_STEP EM maximization step for the mixed binary/continuous learning-curve model

Usage:

[alph, beta, gamma, rho, sig2e, sig2v, xnew, muone] = …

m_step(N, Z, signewsq, xnew, a, muone, startflag, binflag)

Inputs:

N : 1xK double - binary/count observations per trial – required Z : 1xK double - continuous observations (e.g. RT) – required signewsq : 1xK+1 double - smoothed state variance SIG^2{k|K} – required xnew : 1xK+1 double - smoothed state estimate x{k|K} – required a : 1xK+1 double - smoother gain A{k} – required muone : double - logit of background probability (pass-through) – required startflag: integer - initial-condition rule (0: fixed, 2: estimated) – required binflag : logical - if true, suppress RT model (force alpha=beta=0) – required

Outputs:

alph : double - updated RT intercept beta : double - updated RT slope gamma : double - binary observation state weight (fixed to 1) rho : double - state AR(1) coefficient (fixed to 1) sig2e : double - updated RT noise variance sig2v : double - updated state variance xnew : 1xK+1 double - smoothed state (passed through) muone : double - logit of background probability (passed through)

Notes

Implements the closed-form M-step updates using sufficient statistics W{k|K}, W{k-1|K}, W{k,k-1|K}. Sensitive to the initial-condition specification via startflag. gamma and rho are held fixed.

See also: em_bino, mixedlearningcurve, mixedlearningcurve2

∿∿∿ Prerau Laboratory MATLAB Codebase · sleepEEG.org ∿∿∿