Stats_EO Flashcards
Quels sont les pourcentages selon les écarts type d’une distribution normale?
+/- 1 écart type = 68%
1 EC = 34%
+/- 2 écart type= 95%
2 EC = 47.7%
+/- 3 écart type = 99.7%
3 EC = 49.8%
Quelle est la différence entre validité et fiabilité?
Validité: est-ce que les résultats obtenus mesurent ce que l’on voulait mesurer? = applicability ; construct
Fiabilité: reproductibilité des résultats = interrater reliability, test-retest correlation
définition de la p value
Seuil selon lequel le resultat obtenue est expliqué par la chance
Eg P <0.05 = le résultat est expliqué par la chance moins de 5% du temps = seuil pour réfuter l’hypothèse nulle
What’s an alpha error?
Type 1 error
Internal validity error
Null hypothesis is rejected but was true
(i.e your assumption of no effect is incorrect because there is a positive association)
= significance level (p=0.05)
what is beta error?
Type 2 error
Power = ability to detect true effect = 80% (on est plus tolérant que alpha parce que si on trouve pas l’association, quelqu’un d’autre va la trouver) EXAMPLE: grosse bédaine chez un femme et le médecin dit qu’elle n’est pas enceinte)
Null hypothesis should not have been accepted
(I.e. the is an effect present, but we accept the null hypothesis)
What are the qualities of a good screening test?
Cheap Easy to administer Not intrusive Reliable Valid Compared to a gold standard
How to calculate odds ratio?
AD/BC
Def: measures if exposure is associated with higher or lower odds of outcome. Used in restrospective date of case control study. When outcome is rare (eg rate disease), it is a good approximation of risk ratio.
how to calculate risk ratio?
A/(A+B) / C/(C+B)
Def: measure of incidence of an outcome in exposed people compared to outcome in nonexposed people
If RR=1 no association, risk of exposed = risk of non exposed
If RR >1 positive associate, risk of exposed greater than risk in non-exposed
If RR< 1 negative association ; exposed may be protective
What is risk difference?
also called absolute risk reduction ARR
A/ (A+B) - C/(C+D)
How to calculate NNT?
1/ARR
Nbr needed to treated for one to show response
What is the hawthorne effect?
People who know they are being studied modify their behavior and do better than the average patient
What does a sensitive test mean? How to calculate?
How good is a test at picking up people who have the disease
A/A+C
What does specificity mean? How to calculate?
How good is a test at excluding those who don’t have the disease
D/(B+D)
What does positive predictive value mean? How to calculate?
Informs the chance of having a disease is the test results is positive
A/(A+B)
What does negative predictive value mean? How to calculate?
Informs the chance of not having the disease if test is negative
C/(C+D)
