Collateral information

Collateral information
See also

Collateral information are information that are necessary for the full or proper use of basic statement. Very often, this is secondary communiqe, made available as valuable source to maximize the use of owned knowledge[1].


Collateral information is one of the most important factors that has a significant impact on the final result of a given project. Additional sources of information can refer to many different entities: Both family members, suppliers, friends, contractors, competitors, should be bound to any previously made entries. Support guides can provide objective data and improve the rating or support the specific outcome that we care about.

The development of collateral information requires additional time and effort; It can include various types of training or material resources. Consumers may perceive additional time as a delay, and if it is associated with a poor result, the lack of time and the collection of additional data may be perceived by some as an error[2].

Collateral information in IRT

In addition to the regular information in the response data, two sources of collateral information are identified[3]:

  • the joint information in the responses
  • the information from the previous distribution (These one is natural empirical).

Unlike traditional hierarchical item response theory (IRT) modeling, where the gain in estimation accuracy is typically paid, use of this posterior predictive distribution improves both the accuracy and the bias of IRT parameter estimates. In an empirical studies, the improvements are demonstrating for the estimation of the person and item parameters in a three-parameter response[4].


To emphasize that the borrowed information is collected simultaneously with the direct information on the parameters, Novick and Jackson (1974) introduced the notion of collateral information. This term avoids the more temporal connotation. It should be noticed that the use of the term information differs from that elsewhere in scientific endeavors, where it is typically taken to imply that observations can be predicted from other variables[5]. However, collateral information in the hierarchical sense does not require the presence of any predicting variables but is already available if the units of observation can be assumed to follow a common distribution. If the assumption holds, as soon as data are collected for the parameters of some of the units, information is received on all of them, for example, about their typical range of values. Surprisingly, that collateral information is specific for the individual items and persons leads to improvement of both the accuracy[6].


  1. Smit A., Kelderman H., Van der Flier H. (1999), vol. 4 s. 22-23
  2. Carruthers P., Stich S., Siegal M. (2014), s. 171-173
  3. Smit A., Kelderman H., Van der Flier H. (2000) vol. 5 s. 33-37
  4. Wim J., Rinke H., Fox J.P. (2010) s. 328-335
  5. Desbarats A.J., Logan C.E. Hinton M.J., Sharpe D.R. (2002), s. 25-38
  6. Mislevy R.J., Sheehan K.M. (2009), vol.54 s. 661-679


Author: Filip Piszczek