Most scientific findings in many research fields are subject to randomness, since they are, in fact, based on statistical tests that are a function of random samples. Nevertheless, researchers seem to ignore this dependence on randomness, especially when facing reproducibility problems. In this paper, the probability of reproducing significant findings (RP) is held to be the main piece of information to be associated with any finding of scientific and statistical relevance. Basic tech niques for estimating RP are reviewed and it is shown they present wide variability in estimation. RP estimates should be adopted to evaluate the reliability/stability of statistical test outcomes and in most situations RP-testing can substitute the p-value, which is difficult to understand and misleading. Also, the RP perspective allows an intuitive reading of results, making researchers aware of the randomness of scientific findings based on statistical tests.
DE MARTINI, D. (2015). Reproducibility and Probability [Rapporto tecnico].
Reproducibility and Probability
DE MARTINI, DANIELEPrimo
2015
Abstract
Most scientific findings in many research fields are subject to randomness, since they are, in fact, based on statistical tests that are a function of random samples. Nevertheless, researchers seem to ignore this dependence on randomness, especially when facing reproducibility problems. In this paper, the probability of reproducing significant findings (RP) is held to be the main piece of information to be associated with any finding of scientific and statistical relevance. Basic tech niques for estimating RP are reviewed and it is shown they present wide variability in estimation. RP estimates should be adopted to evaluate the reliability/stability of statistical test outcomes and in most situations RP-testing can substitute the p-value, which is difficult to understand and misleading. Also, the RP perspective allows an intuitive reading of results, making researchers aware of the randomness of scientific findings based on statistical tests.File | Dimensione | Formato | |
---|---|---|---|
main_document.pdf
Solo gestori archivio
Dimensione
220.39 kB
Formato
Adobe PDF
|
220.39 kB | Adobe PDF | Visualizza/Apri Richiedi una copia |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.