On tests and significance in econometrics

Keuzenkamp, Hugo A.; Magnus, Jan R.

doi:10.1016/0304-4076(94)01624-9

Cited by 54 publications

(34 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Actually, scientists often avoid that conclusion (cf. Keuzenkamp and Magnus 1995). One reason for this reluctancy is certainly model uncertainty.…”

Section: Summary and Discussionmentioning

confidence: 99%

Science without (parametric) models: the case of bootstrap resampling

Sprenger

2009

Synthese

View full text Add to dashboard Cite

Scientific and statistical inferences build heavily on explicit, parametric models, and often with good reasons. However, the limited scope of parametric models and the increasing complexity of the studied systems in modern science raise the risk of model misspecification. Therefore, I examine alternative, data-based inference techniques, such as bootstrap resampling. I argue that their neglect in the philosophical literature is unjustified: they suit some contexts of inquiry much better and use a more direct approach to scientific inference. Moreover, they make more parsimonious assumptions and often replace theoretical understanding and knowledge about mechanisms by careful experimental design. Thus, it is worthwhile to study in detail how nonparametric models serve as inferential engines in science.Keywords: models, data, inductive inference, nonparametric statistics, bootstrap resampling Probabilistic ModelingModeling plays a key role in empirical science, especially when overarching theories cannot be applied. Many efforts in science focus on constructing, comparing and revising models of physical entities, phenomena and processes. Bohr's model of the atom, Volterra's model of predator-prey populations and the random walk model for the motion of molecules in a fluid are among the most popular ones. Models enable us to recognize fundamental relations between physical quantities, to understand the effects of causal interventions and to generalize observed effects to more complex and realistic cases. Often, their construction is triggered by concrete puzzles: For instance, Volterra (1926) developed his mathematical model of predator-prey population dynamics in response to the surprising shortage of adriatic fish after World War I. Volterra's model started from abstract considerations, but its predictions were found to be in stunning agreement with reality (see Weisberg's (2007) case study for more details). The way the Volterra model has been developed, refined and transferred to other scientific inquiries exemplifies a general strategy: to set up mathematically tractable models which capture fundamental mechanisms of the underlying system, and to gradually amend and refine them in order to account for the complexity of large-scale systems in the real world. In other words, models allow us to discover characteristic regularities (e.g. cycles in the population dynamics) as well as to explain concrete phenomena, such as "why does a disruption in fishing activity increase the predator/prey ratio?".Hence, it is not surprising that philosophers of science have been spending a lot of paper on the various features of model-building. In particular, they studied † Contact information:

show abstract

“…Actually, scientists often avoid that conclusion (cf. Keuzenkamp and Magnus 1995). One reason for this reluctancy is certainly model uncertainty.…”

Section: Summary and Discussionmentioning

confidence: 99%

Science without (parametric) models: the case of bootstrap resampling

Sprenger

2009

Synthese

View full text Add to dashboard Cite

show abstract

“…It should also be noted that these inferential methods are not intended for large or massive samples. Keuzenkamp and Magnus (; p.17) point out that Fisher's theory of significance testing is intended for small samples, stating that ‘Fisher does not discuss what the appropriate significance levels are for large samples’, a point also made by Lindley (; section 14.4). Meanwhile, the values of

α^{*}

(optimal significance level discussed in Section 3.4) and

α^{+}

(adaptive significance level given in ) quickly approach zero as the sample size increases, providing a useful guard against the problem.…”

Section: A Statistical Toolboxmentioning

confidence: 99%

Tackling False Positives in Business Research: A Statistical Toolbox With Applications

Kim

2018

Journal of Economic Surveys

View full text Add to dashboard Cite

Serious concerns have been raised that false positive findings are widespread in empirical research in business disciplines. This is largely because researchers almost exclusively adopt the ‘p‐value less than 0.05’ criterion for statistical significance; and they are often not fully aware of large‐sample biases which can potentially mislead their research outcomes. This paper proposes that a statistical toolbox (rather than a single hammer) be used in empirical research, which offers researchers a range of statistical instruments, including a range of alternatives to the p‐value criterion such as the Bayesian methods, optimal significance level, sample size selection, equivalence testing and exploratory data analyses. It is found that the positive results obtained under the p‐value criterion cannot stand, when the toolbox is applied to three notable studies in finance.

show abstract

“…Summers (1991) criticises this method, suggesting that without some idea of the power of statistical tests against interesting alternative hypothesis and/or some metric for evaluating the extent to which the data is consistent with a maintained hypothesis, formal statistical tests are uninformative. From early application to most recent work investigators have relied on 1% and 5% levels of significance (Keuzenkamp & Magnus, 1995). Testing the validity of a hypothesis using these fairly arbitrary levels of significance was opposed by Pearson.…”

Section: The Traditional Approachmentioning

confidence: 99%

Econometric Methodology I: The Role of Theory and Observation in Agricultural Economics Research / Ekonometriese Metodologiei: Die Rol Van Teorie en Waarneming in Landbou-Ekonomiese Navorsing

Townsend

1997

Agrekon

View full text Add to dashboard Cite

On tests and significance in econometrics

Cited by 54 publications

References 26 publications

Science without (parametric) models: the case of bootstrap resampling

Science without (parametric) models: the case of bootstrap resampling

Tackling False Positives in Business Research: A Statistical Toolbox With Applications

Econometric Methodology I: The Role of Theory and Observation in Agricultural Economics Research / Ekonometriese Metodologiei: Die Rol Van Teorie en Waarneming in Landbou-Ekonomiese Navorsing

Contact Info

Product

Resources

About