Conditional estimation of Rasch model

estimate() estimates the parameters of the Rasch/Bradley-Terry_Luce model with a conditional maximum likelihood procedure. This is an iterative procedure using Newton's method of optimization. The algorithm concept was initially developed by Rasch (1960) and implemented by Pollitt (2012) and NoMoreMarking ltd. among others.

Our module was originally based on the code of NoMoreMarking ltd. and has been further optimised.

Tests

You can run the unit tests with

$ yarn test

API

Item

Type: Object

Properties

• id string ID of the item
• ranked boolean whether or not the item is ranked. Only unranked items (i.e. ranked: false) are estimated
• ability number the ability
• se number standard error

Comparison

Type: Object

Properties

• a string the ID of the "A" item
• b string the ID of the "B" item
• selected string?? the ID of the selected item

estimate

Estimates the items featured in comparisons

Parameters

• payload Object
  • payload.comparisons Array<Comparison>
  • payload.items (Array<Item> | {}) An Array or a map of Item

Returns (Array<Item> | {}) An Array or a map of Item (reflects the type of parameter items)