Lecture 7 Flashcards
Power
Video 1
What does power mean?
The probability to detect an effect that is actually there. Underpowered studies are bound to fail
Video 2
What is a type I error?
When you reject the null hypothesis, but it is actually true
Video 2
What is a type II error?
When you falsly accept the null hypothesis
Video 2
Do you set the beta as well?
No, you only set the alpha, the beta follows
Video 2
What is sensitivity?
The true positives rate = true positive / all positive
Video 2
What is specificity?
The true negative rate
Video 3
What is the effect of a high effect size?
There is less overlap in the distributions, gives a higher power
Video 3
What is the result of a high sample size?
There is less variation in the sampling, a more accurate t, distributions in the same position but are more pieked. Gives a higher power
Video 3
How is the total Var calculated?
Var manipulation (control through it’s severity) + Var random nois (controol via careful matching/selection) + Var measurement error (control good intrument)
Video 3
What is a good instrument?
One that is standardized, reliable, objective, has a good resolution
Video 3
What is resolution?
The variation of the population is not taken into account, skewed. Is also known as censoring. Answer everyhting correc: ceiling effect. Answer all wrong: bottom effect
Video 3
When is resolution lost?
When you make data dichotomous
Video 4
What is the consequence of low power?
There is a higher chance of type II errors.
Video 4
What is the Winner’s curse?
The effect of a low powered study is overestimated/inflated. When the effect is close to 0.05, other studies will probably get a higher p
Video 4
When is a low power study unethical?
When animals are used and there is a low probability of succeeding
Video 4
What is the Positive predictive value (PPV)?
The true postives/ all postives
Video 4
What is a safe study?
It has a high prevalence of effect/ disease
Video 4
What is a risky study?
A low prevalance of the disease (if it is well powered, the results are more credible)
Video 4
What does the PPV indicate?
The credibility of the results: how likely it is that the effect found is a real effect
Video 4
What is a different formula for calculating PPV?
power * odds / (power * odds + alpha)
Video 4
How can the odds be interpreted?
When the odds are 1/9, 1 effect to 9 non-effects
Video 5
What is a priori power?
Want to know whether you have sufficient power before you do the study
