Bayesian estimation of a multilevel IRT model using gibbs sampling

In this article, a two-level regression model is imposed on the ability parameters in an item response theory (IRT) model. The advantage of using latent rather than observed scores as dependent variables of a multilevel model is that it offers the possibility of separating the influence of item difficulty and ability level and modeling response variation and measurement error. Another advantage is that, contrary to observed scores, latent scores are test-independent, which offers the possibility of using results from different tests in one analysis where the parameters of the IRT model and the multilevel model can be concurrently estimated. The two-parameter normal ogive model is used for the IRT measurement model. It will be shown that the parameters of the two-parameter normal ogive model and the multilevel model can be estimated in a Bayesian framework using Gibbs sampling. Examples using simulated and real data are given.     

The role of learning in risk-avoidance strategies during spider–ant interactions

Cognitive abilities used by arthropods, particularly predators, when interacting in a natural context have been poorly studied. Two neotropical sympatric predators, the golden silk spider Nephila clavipes and the ectatommine ant Ectatomma tuberculatum, were observed in field conditions where their interactions occurred regularly due to the exploitation of the same patches of vegetation. Repeated presentations of E. tuberculatum workers ensnared in their web triggered a progressive decrease in the capture response of N. clavipes. All the spiders that stopped trying to catch the ant on the second and/or third trial were individuals that had been bitten during a previous trial. Behavioural tests in natural field conditions showed that after a single confrontation with ant biting, spiders were able to discriminate this kind of prey more quickly from a defenceless prey (fruit flies) and to selectively and completely suppress their catching response. This one-trial aversive learning was still effective after 24 h. Likewise, E. tuberculatum workers entangled once on a N. clavipes web and having succeeded in escaping, learned to escape more quickly, breaking through the web by preferentially cutting spiral threads (sticky traps) rather than radial threads (stronger structural unsticky components) or pursuing the cutting of radials but doing it more quickly. Both strategies, based on a one-trial learning capability, obviously minimize the number of physical encounters between the two powerful opponents and may enhance their fitness by diminishing the risk of potential injuries resulting from predatory interactions.     

Algorithmic complexity for short binary strings applied to psychology: a primer

As human randomness production has come to be more closely studied and used to assess executive functions (especially inhibition), many normative measures for assessing the degree to which a sequence is randomlike have been suggested. However, each of these measures focuses on one feature of randomness, leading researchers to have to use multiple measures. Although algorithmic complexity has been suggested as a means for overcoming this inconvenience, it has never been used, because standard Kolmogorov complexity is inapplicable to short strings (e.g., of length l?≤?50), due to both computational and theoretical limitations. Here, we describe a novel technique (the coding theorem method) based on the calculation of a universal distribution, which yields an objective and universal measure of algorithmic complexity for short strings that approximates Kolmogorov–Chaitin complexity.