Test choice
3 important questions on Test choice
What are the misuses of p-value?
- Large p-value means no difference: wrong
- p-value decreases when sample size increases
- absence of evidence is not evidence of absence
- Multiple testing & 0.05
- doing experiment 100 times and finally succeed, can 0,05 be used
- 1 dataset: many models to test which hypothesis fits best, can 0.05 be used? Type I error
- Multiple testing: Bonferroni correction for example
- Smaller p-value is more significant? Not necessarily
- effect size is also important
How do you add more than 1 independent variable?
- First: 1 by 1 and look at the effect
- this is often not final result: high change of Type I error
- Then variable together: Forward and Backward
- forward: start with 1 and add 1 by 1
- backward: start with all and drop 1 by 1
- stop when model has only significant variables
What do you base you selection of a model on?
- Hypothesis
- p-values for independent variables
- effect size
- adjusted R^2
- information criteria: AIC
The question on the page originate from the summary of the following study material:
- A unique study and practice tool
- Never study anything twice again
- Get the grades you hope for
- 100% sure, 100% understanding
Remember faster, study better. Scientifically proven.
