A Note on the Coefficient of Determination in Models with Infinite Variance Variables

Kurz‐Kim, Jeong‐Ryeol; Loretan, Mico

doi:10.2139/ssrn.996664

“…Statistical analysis technique, co-efficient of determination (R 2 ) was used for methodology validation. It determines how close the predicted data matches the original data, according to the equation ( 9) [27].…”

Section: Case Studymentioning

confidence: 99%

Artificial Neural Networks Supporting Cause-and-Effect Studies in Product–Service System Development

Aeddula

¹

,

Wall

²

,

Larsson

³

2021

Design for Tomorrow—Volume 2

1

0

View full text Add to dashboard Cite

A data analysis method based on artificial neural networks aiming to support cause-and-effect analysis in design exploration studies is presented. The method clusters and aggregates the effects of multiple design variables based on the structural hierarchy of the evaluated system. The proposed method is exemplified in a case study showing that the predictive capability of the created, clustered, dataset is comparable to the original, unmodified, one. The proposed method is evaluated using coefficient-of-determination, root mean square error, average relative error, and mean square error. Data analysis approach with artificial neural networks is believed to significantly improve the comprehensibility of the evaluated cause-and-effect relationships studying PSS concepts in a cross-functional team and thereby assisting the difficult and resource-demanding negotiations process at the conceptual stage of the design.

show abstract

“…Statistical analysis technique, co-efficient of determination (R 2 ) was used for methodology validation. It determines how close the predicted data matches the original data, according to the equation ( 9) [27].…”

Section: Case Studymentioning

confidence: 99%

Design for Tomorrow—Volume 1

2021

Smart Innovation, Systems and Technologies

3

0

View full text Add to dashboard Cite

A data analysis method based on artificial neural networks aiming to support cause-and-effect analysis in design exploration studies is presented. The method clusters and aggregates the effects of multiple design variables based on the structural hierarchy of the evaluated system. The proposed method is exemplified in a case study showing that the predictive capability of the created, clustered, dataset is comparable to the original, unmodified, one. The proposed method is evaluated using coefficient-of-determination, root mean square error, average relative error, and mean square error. Data analysis approach with artificial neural networks is believed to significantly improve the comprehensibility of the evaluated cause-and-effect relationships studying PSS concepts in a cross-functional team and thereby assisting the difficult and resource-demanding negotiations process at the conceptual stage of the design.

show abstract

“…On a related note,Kurz-Kim and Loretan (2007) show that, in the case of the well-known finite sample F-distribution of the R 2 under the null that the latter is zero, this might be a concrete danger when the normality assumption fails and the regression variables have fat tails distributions, as it is often the case for regressions involving currency returns.…”

mentioning

confidence: 99%

Predictability and ‘good deals’ in currency markets

Levich

¹

,

Potì

²

2015

International Journal of Forecasting

View full text Add to dashboard Cite

Comments welcomedThis paper studies predictability of currency returns over the period . To assess the economic significance of predictability, we construct an upper bound on the explanatory power of predictive regressions. The upper bound is motivated by "no good-deal" restrictions that rule out unduly attractive investment opportunities. We find evidence that predictability often exceeds this bound. Excess-predictability is highest in the 1970s and tends to decrease over time, but it is still present in the final part of the sample period. Moreover, periods of high and low predictability tend to alternate. These stylized facts pose a serious challenge to Fama's (1970) Predictability and 'Good Deals' in Currency MarketsThis paper studies predictability of currency returns over the period . To assess the economic significance of predictability, we construct an upper bound on the explanatory power of predictive regressions. The upper bound is motivated by "no good-deal" restrictions that rule out unduly attractive investment opportunities. We find evidence that predictability often exceeds this bound. Excess-predictability is highest in the 1970s and tends to decrease over time, but it is still present in the final part of the sample period. Moreover, periods of high and low predictability tend to alternate. These stylized facts pose a serious challenge to Fama's (1970) Efficient Market Hypothesis but are consistent with Lo's (2004) Adaptive Market Hypothesis, coupled with slow convergence towards efficient markets. Strategies that attempt to exploit excess-predictability are very sensitive to transaction costs but those that exploit monthly predictability remain attractive even after realistic levels of transaction costs are taken into account and are not spanned either by the Fama and French (1993) equity-based factors or by the AFX Currency Management Index.

show abstract

A Note on the Coefficient of Determination in Models with Infinite Variance Variables

Cited by 5 publications

References 31 publications

Artificial Neural Networks Supporting Cause-and-Effect Studies in Product–Service System Development

Artificial Neural Networks Supporting Cause-and-Effect Studies in Product–Service System Development

Design for Tomorrow—Volume 1

Predictability and ‘good deals’ in currency markets

Contact Info

Product

Resources

About