Theil's Forecast Accuracy Coefficient: A Clarification

Bliemel, Friedhelm

doi:10.1177/002224377301000413

Cited by 116 publications

(36 citation statements)

References 1 publication

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…U 2 = 0 indicates a perfect fit between the two data sets. U 2 = 1 is observed 'when the prediction method is naive no-change extrapolation or when it leads to the same standard deviation of forecast error as that method' (Bliemel, 1973).…”

Section: Error-based Measuresmentioning

confidence: 96%

See 1 more Smart Citation

A review of performance criteria to validate simulation models

Hora

Campos

2015

Expert Systems

View full text Add to dashboard Cite

This study reviews performance criteria adequate to validate simulation models through the comparison of two quantitative data sets, concerning historical and simulated data. The criteria reviewed were organized according to its characteristics into the groups: error‐based measures, information theory measures, information criteria, parametric tests, non‐parametric tests, distance‐based measures and combined measures. Each criterion is reviewed through its mathematic definition, its applications in literature and the identification of its advantages and drawbacks. The features assessed by each criterion are identified and discussed. This study provides a concise outline over the criteria reviewed, which can be used as a guide to help developers of simulation models into the decision on the most appropriate criteria to validate their models.

show abstract

Section: Error-based Measuresmentioning

confidence: 96%

“…When U 1 = 0, the estimation is completely coincident with the observations (x i = y i , ∀ i), indicating a perfect forecast. The case U 1 = 1 indicates the maximum inequality (when there is negative proportionality over the two data sets or when one of the data sets is identical to zero) (Theil, 1966;Bliemel, 1973).…”

Section: Error-based Measuresmentioning

confidence: 99%

A review of performance criteria to validate simulation models

Hora

Campos

2015

Expert Systems

View full text Add to dashboard Cite

show abstract

“…23) and UII (Theil, 1965;p.28) are suggested in the literature. An interpretation of the two measures is found in Bliemel (1973) who concludes that UII gives more information about the forecasting accuracy. Therefore, in this paper, we use UII (denoted simply by U below) which takes the following expression for a particular model:…”

Section: Comparison With Alternative Model Specificationsmentioning

confidence: 99%

On the superiority of pulsing under a concave advertising market potential function

Mesak

Ellis

2009

European Journal of Operational Research

View full text Add to dashboard Cite

“…It shows the forecast errors for the Box-Jenkins model and the exponential smoothing model compared with the yearly reinitialization atypical data modified forecast. As a measure of accuracy, Theil's U statistic is used (Bleimel 1973). The U value drops from .103 to .075.…”

Section: Replication Of the Short-time Senesmentioning

confidence: 99%