TOC

TOC

The classical test theory has limitations as a principle by which to guide a test constructor. Latent Trait Theory has several advantages over classic test theory. For one, using latent test theory, an examinee's ability can be estimated independently of the particular choice on the number of items. This is a considerable advantage, for examinees can be compared though they have taken different subsets of tests or totally different tests. The advantages of these are far reaching.
Unlike in classical test theory, each examinee's score has essentially its own estimate of the error, where it is the same for all the scores in a particular test regardless of one's ability i.e. ones score. The item parameters of latent trait models such as the examinee's ability, item difficulty, item discrimination, etc, are independent of the groups or sub-groups(population). They are universal, that is, they carry universal meaning. Thus meaningful comparison is possible across different populations of interest. This being the case, latent trait model parameters are useful in development of item banks and in detecting item bias. There are disadvantages with the latent trait theory. For one it is a highly complex theory touching on multivariate statistics from test theory. Latent trait models will need to be postulated, a reseracher being guided by substantive heory and practice. Computer programs need to be used, above all to estimate the parameters of the model. This adds to the already complex area of latent trait theory. The paper has attempted to tackle this subject from a descriptive point of view, avoiding mathematics as much as possible.