Lecture 9 Flashcards
Critical perspectives on statistics and scientific literature
Video
What is the probability of a type 1 error if H0 is true?
0.05 (or 5%)
Video
What is stronger, a confirmed theoretical relationship or an observed potential relationship?
A confirmed theoretical relationship, it has a lower chance of type 1 error
Video
What is hypothesis free testing?
When a test or research is done that is not driven by a theory
Video
What does the PH0 depend on?
Depends on the replicability of the findinf and the theoretical basisi of the hypothesis.
Video
What is the risk of explorative research
A high risk of type 1 error
Video
What is the base rate?
The prevelance of a disease
Video
What is the formula for sensitivity?
P (pos. diagnosis I disease)
Video
What is the formula for specificity?
P (neg. diagnosis I disease)
Video
What is the formula for the base rate?
P (disease)
Video
What causes the base rate to be high?
A high PPV or a low NPV
Video
What can a quality test be influence by?
The PPV and the NPV
Video
What are the equivalents of sensitivity and specificity in statistics?
Sensitivity = power
Specificity = confidence
Video
What do you calculate when calculating the base rate in statistics?
The probability that H0 is false
Video
What is base rate falicy?
The fact that the base rate is mostly overlooked in statistics
Video
What is the min. base rate that you need and why?
The minimum base rate is 70 % to keep a type 1 error within the bounds of 5%
Video
When do you have a low chance of type 1 error?
When there is a high base rate, it is replicated (with a different methodology as well), generate a H0 and seek confirmation
Video
What type of bias does the publication bias lead to?
Leads to knowledge bias
Video
What is Proteus the symbol for?
For new/large insignificant effects
Video
What is the effect of multiple replications of the same study?
The meta-analysis will stabilize over time and will show the true effect size
Video
What is different about replication studies?
They are hard to publish
Video
What is the general problem with small/ non-randomized studies?
They show non-replication
Video
Where can you find publication biases?
In FDA
Notes
What is Russell’s turkey theory?
It is based on observations, they can change there is no certainty
Notes
When is exploratory research better?
When there is replication and there are theoretical foundations
