Biometry Hub Seminar
- Date: Tue, 25 May 2021, 10:00 am - 11:00 am
- Location: Online Zoom meeting
- Contact: Beata Sznajder firstname.lastname@example.org
- Dr Murthy Mittinty (University of Adelaide) Presenter
Please join us for the Biometry Hub seminar where Dr Murthy Mittinty of the School of Public Health of the University of Adelaide will present on a new procedure for quantitative assessment of yield/quality improvement – trend direction assessment.
Please also note that this seminar will be recorded. If you have a question to the speaker but would rather not be recorded, please send question via chat during the meeting and Beata Sznajder will ask it on your behalf. Please email Beata Sznajder for details of the Zoom meeting.
Has yield/quality improved or been maintained? A quantitative assessment procedure
Dr Murthy Mittinty (School of Public Health, University of Adelaide)
Many policies require reporting on trend in yields/quality. This is usually addressed by testing a hypothesis positing that there was zero slope in some parameter of the sample population a given period. Failure to achieve “statistical significance” is often falsely interpreted as evidence that there was no trend of concern—the P-value of these tests can become ever smaller as the sample size increases and so can the detectable trend. To avoid this problem, a new trend direction assessment (TDA) procedure is proposed, based on a formulation proposed by Jones and Tukey (2000) that considers error risks when inferring the direction of difference in population means. Abandoning the null hypothesis testing procedure the TDA approach allows calculate probabilities that inform if the yield/quality has been increasing or decreasing. Using this continuous distribution of the probabilities one can summarise the trend as “extremely likely” or “unlikely” compared to using uninterpretable terminology such as “Statistically not significant”. This trend assessment procedure requires no additional information than a traditional test, for which the level of significance is reinterpreted as misclassification error rate (inferring a decrease when in fact there was an increase or vis-a-vis). Some applications will be shown to demonstrate the procedure. This procedure also possess a possible frame work that addresses the more complex question of whether the yield has been “maintained”, in which trend magnitude of policy significance must be defined.