מדע במשבר (0431.3870)
שם הקורס באנגלית: Science in crisis
can we trust the results of our experiments and statistical tests?
מרצה הקורס: פרופ' שי מאירי
ימים ושעות הקורס
| סוג: שיעור | ||||||
| מספר קורס | סמסטר | יום | משעה | עד שעה | בניין | חדר |
| 0431.3870.01 | א | ב | 11:00 | 13:00 | שרמן | 330 |
סילבוס:
A graduate level course
(inspired by: https://hardsci.wordpress.com/2016/08/11/everything-is-fucked-the-syllabus/)
Topics:
- Science in crisis – replicability issues, and the problem of power and generality with small samples
- Statistics in crisis: Issues with probabilistic statistics
- Meta-analysis to the rescue
- Meta-analysis in crisis – the file drawer problem, publication bias, and publishing practice
- Types of science: experimental, observational and theoretical, how hard is hard science?
- Big data to the rescue: macroecology and evidence based science
- Interlude: bad science, non-science and pseudo-science (we may be bad, but there are worse)
- Big data in crisis – low variances and issues with model selection (the world views of AIC, BIC, model averaging and junk explanatory variables)
- Potential remedies and overview
- Structure: We will start with an overview by the lecturer, with introduction to power and meta-analyses and the file-drawer problem. Students will then choose subjects from the list above at the beginning of the semester, read about them and present them in class (~45 minute presentations), to be followed by discussion.
- Pre-requisites: Biostatistics course and one more piece of proven record in either learning statistics (advanced level courses such as the R course etc.) or using them (submitting a paper for publication or submitting a thesis where statistics were used)
- Semester: Aleph
- Credit: 3 points
- Final grade:
- Presentation (30%)
- intelligent participation in the discussion (20%)
- house exam (50%, basically a review of the subject presented to the class)
-
A Reading list will include papers from the primary literature. The students are required to read relevant papers for preparing their presentation (at least 4 papers per topic, detailed list will be provided).
