2022
DOI: 10.1126/sciadv.abn9450
|View full text |Cite
|
Sign up to set email alerts
|

Models with higher effective dimensions tend to produce more uncertain estimates

Abstract: Mathematical models are getting increasingly detailed to better predict phenomena or gain more accurate insights into the dynamics of a system of interest, even when there are no validation or training data available. Here, we show through ANOVA and statistical theory that this practice promotes fuzzier estimates because it generally increases the model’s effective dimensions, i.e., the number of influential parameters and the weight of high-order interactions. By tracking the evolution of the effective dimens… Show more

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
3
1
1

Citation Types

0
21
0

Year Published

2023
2023
2024
2024

Publication Types

Select...
8
1
1

Relationship

1
9

Authors

Journals

citations
Cited by 41 publications
(21 citation statements)
references
References 59 publications
0
21
0
Order By: Relevance
“…Some studies have used multi-parametric approaches, while we use a simple relation to a single predictor. Complex mathematical approaches do not necessarily provide more reliable results, and, in fact, potentially produce more uncertain estimates 74 . Admittedly, it is unrealistic to model fluvial pCO 2 using a single predictor at continental or global scales.…”
Section: Resultsmentioning
confidence: 99%
“…Some studies have used multi-parametric approaches, while we use a simple relation to a single predictor. Complex mathematical approaches do not necessarily provide more reliable results, and, in fact, potentially produce more uncertain estimates 74 . Admittedly, it is unrealistic to model fluvial pCO 2 using a single predictor at continental or global scales.…”
Section: Resultsmentioning
confidence: 99%
“…In modelling, when there are available data against which to compare the model predictions, information criterion such as Akaike’s ( 2011 ) or Schwarz’s ( 1978 ) can be used to balance model complexity with parsimony. Lacking a validation data set, uncertainty quantification and SA can be used to gauge the uncertainty in the inference and its sources (Puy et al, 2022 ) For each family of quantification, agreed rules should be established to gauge complexity. Consider if the degree of complexity of a quantification can be gauged against agreed criteria.…”
Section: Bridging Mathematical Modelling With Sociology Of Quantifica...mentioning
confidence: 99%
“…This formulation possesses important significance for random systems without governing equations and excitation data. On the other hand, a low-dimensional description means higher certainty than the original high-dimensional description [29]. One work, and the only one work, in regard to this subject, comes from the same authors [30].…”
Section: Introductionmentioning
confidence: 99%