item.selection: Item Selection
In changxiulee/BayesianRasch: Bayesian Rasch Modeling

Given that a set of item is not unidimensional, this function helps to determine which item should be removed. To do this, user need to first select a small set of item (core.item) that is known for sure to be unidimenional. This small set of item will be subject to Rasch PCA to verify this belief. Then, for each item outside core.item (peripheral item), a Rasch Analysis will be conducted together with the core item and the item fit is computed. If an item does not share a common dimensional with the core item, the item fit of the peripheral item is worse (> 1.3). Please note that, in this function, all Rasch model is estimate using ltm / MMLE, not Bayes because 1) estimating using Bayes is too slow and not suitable when we want to estimate a model repeatedly 2) We do not utilized the uncertainty information in this function, 3) LTM's MMLE is identical to Bayesian's Maximum a Posterior most of the time and it is faster.

1	item.selection(data, core.item, peripheral.item)

`data`	A data frame containing the data
`core.item`	A set of item that is obviously measuring the dimension in question
`peripheral.item`	Item outside of core item