Fichas
statistical tests A
Fichas del conjunto (13)
- 1Internal ValidityFor population examined, is observed affect really a consequence of the treatment?
- 2Threats of Internal Validity (1)1. Confounding in observational studies
-> Solution: Experimental studies that 2 groups only differ in whether they received the treatment - 3Threats of Internal Validity (2)2. Selection Bias
i.e. the selection of treatment group may be such that intervention more/less likely to work, compared to control group
->Solution: Randomisation in allocation of subjects to control & treatment groups - 4Threats of Internal Validity (3)3. Experimental Effects
-> Placebo effect (subjects in treatment group perform better because they know they are in the group)
->Experimenters biased due to prior expectations
Solution: Blinding ->subject & experimenters don't know which group in - 5External ValidityDo conclusions from the study generalise to other situations/wider pop?
- 6Threats to External Validity (1)1. Non-representative sample
-> sample studies may be different from wider pop to which treatment can be applied to - 7Threats to External Validity (2)2. Equilibrium Effects (i.e. Unintended Consequences)
-> possible that wide adoption of treatment (that was successful in small study)
causes it to be less effective - 8Critical regionOutcomes as far as possible from 3 heads (say 0-6 heads) and hv altogether prob of occurring of at most 5%
- 9Significance Level-Probability of false rejections across experiments
P( reject H0 | H0 is true) - 10p-value-shows how strong the evidence is against H0
-lowest probability at which we cant reject H0
[if p-value < 5%, reject H0 at 5% sig. level] - 11One sided tests-Easier to reject H0 (because p-val smaller)
-Artificially inflates probability of false rejection
-exaggerate statistical significance of results - 12Multiple Comparisons-It is when too many hypothesis are tested, and we are bound to find something significant
- 13How to deal with hypothesis test problems?1. Replicate the results with new data
2. Limit number of hypothesis tested
3. Be transparent in journals published regarding decisions taken during research
Información
- Descripción
- Fichas sobre statistical tests A, creado por avery pumpkin el 11/01/2019.
- Creado por
- avery pumpkin
- Visitas
- 1
- Valoración