Estimating a hidden Bernoulli parameter by sequential Bayesian analysis (Q5958654)

A tetanically-stimulated (TS) neuron is said to have failed to fire if its voltage-clamped excitatory postsynaptic current (EPSC) measurement is devoid of a long-term potentiation (LTP) response. This paper provides a method for evaluating the posterior probability of ``failure'' for TS neurons. A sequential Bayes algorithm is employed on an imperfect Bernoulli trial model in order to refine the posterior density of the failure parameter with each EPSC data record processed. The algorithm is applied to simulated EPSC data with TS elicited LTP responses and is shown to coincide very well with the expected presynaptically-induced LTP failure rate observed in vitro.