Premium
Judgmental forecasting of univariate time series
Author(s) -
Harvey Nigel
Publication year - 1988
Publication title -
journal of behavioral decision making
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.136
H-Index - 76
eISSN - 1099-0771
pISSN - 0894-3257
DOI - 10.1002/bdm.3960010204
Subject(s) - univariate , sequence (biology) , representation (politics) , series (stratigraphy) , sort , uncorrelated , computer science , order (exchange) , statistics , time sequence , identification (biology) , econometrics , artificial intelligence , mathematics , machine learning , multivariate statistics , information retrieval , economics , paleontology , genetics , botany , finance , politics , political science , law , biology
To forecast numbers appearing in sequence, do people just take some sort of average of past items or do they use temporal pattern information that they have extracted from the sequence? In an experiment, subjects forecast successive numbers generated by a first‐order auto regressive algorithm. Afterwards, they were asked to generate their own series of numbers to simulate the sequence they had been forecasting. It was found that (1) generation performance was good—subjects acquired an internal representation of the pattern in the sequence while forecasting it; (2) generation and forecasting performance were uncorrelated—this internal representation was not used for forecasting; (3) learnt ability to forecast a sequence did not transfer to another sequence that was of the same type but that subjects believed to come from another source; (4) subjects were good at estimating the probability that their forecasts would be correct but this ability declined with practice.