Third, with small sample size we have to be prepared to make stronger model- ling assumptions. For example, Altman illustrates the use of a parametric test (ANOVA) to compare 3 groups with 8, 9, and 5 patients in his seminal text “Practical statistics for medical research”. 8 With larger samples, we would more readily switch to a non-parametric test such as a Kruskal–Wallis test. With small sample size, we may have to assume linearity of a continuous predictor (Chap. 9) and no interaction between predictors (Chap. 13). We will subsequently have lim- ited power to test deviations from these model assumptions. It hence becomes more important what our starting point of the analysis is. From a Bayesian viewpoint, we could say that our prior information becomes more important, since the information contributed by our study is limited
If you want to change selection, open document below and click on "Move attachment"
- (no access) - BOOK_steyerberg_Clinical prediction models.pdf, p31
|status||not read|| ||reprioritisations|
|last reprioritisation on|| ||suggested re-reading day|
|started reading on|| ||finished reading on|