Research methods Flashcards
What is a type I error?
Finding an association that does not exist.
What is a type II error?
Finding no association when an association actually does exist.
What is colinearity?
When two explanatory variables are highly correlated
What is colinearity?
When two explanatory variables are highly correlated
What are confounding variables?
Two explanatory variables which are both related to the outcome of each other.
What is an explanatory variable?
Independent variable.
What is an observational study? What are the three types of observational study?
Observe exposure and outcome in different groups.
- Cohort study= following X and Y groups until something happens.
- Cross sectional study=collecting data all from one time point.
- Case control study= retrospectively analysing what one group was exposed to more than the control.
What is an experimental study? What is one benefit of using an experimental study?
Manipulating variables of interest and observing the outcome.
It allows us to see causality, this can never be seen with observational studies.
Statistically speaking, why should we use replication in our studies?
A larger sample size/ study being done more times and more data being collected increases the power to detect important increases/ decreases/changes.
What is the power value we use that is equivalent to a 95% confidence interval?
0.8
What does a power value of 0.8 mean?/ How can we interpret this?
We are 80% sure that there has been no type II error (finding no association when there actually is an association).
What is done in a retrospective study?
Comparing control groups with people with a group of people with an outcome of interest.
What questions do we ask to ensure a study has validity?
Pre study
- Why this location
- Why this (sub)population
- Does the study design make sense
Mid study
- Why this end point (e.g. DALY over morbidity)
- Why this sample size
- Why this duration
Post study
- Why this method of displaying results
- Why this stats test
- Why this conclusion
What are the three qualities we want in our study to make it a good study?
Validity, importance and generalisability
Why do we want a generalisable study?
Allows us to inform different systems, don’t want to to be too specific for only one location/population etc.