| Toward Improved Statistical Practices in Astronomical Studies |
| Eric D. Feigelson |
|
Penn State University & AAS Journals
|
| Astronomical studies face a wide range of data and science analysis challenges that can benefit from modern, often unfamiliar statistical methodology. This talk highlights some common statistical problems and limitations in recent astronomical studies, encouraging use of improved methodology. Topics include: misuse of the Kolmogorov-Smirnov test; limitations in nonlinear regression; over-dependence on nonparametric clustering; inadequate use of spatial statistics and stochastic time series modeling; and guidelines for Bayesian modeling and machine learning studies. |