PARAMETER ESTIMATION METHOD, SYSTEM, AND PROGRAM IN ADJACENT CATEGORY LOGIT MODEL

摘要

PROBLEM TO BE SOLVED: To provide an estimation method, a system, and a program for performing precise and stable parameter estimation for an adjacent category logit model even when the number of samples is low.

SOLUTION: A parameter estimation part 23 estimates a parameter (shown in the figure) maximizing a logarithmic likelihood function lp(θ) with a penalty to data given from an input device 1 to send the estimated parameter to a posterior probability calculation part 22. The posterior probability calculation part calculates a posterior probability of yiby using the parameter (shown in the figure) given from the parameter estimation part for the data given by the input device and sends the calculated posterior probability to an output device 3.

COPYRIGHT: (C)2006,JPO&NCIPI