Medical Statistics Flashcards
What is ontology?
The study of being, theory of knowledge, concerned with what is true.
What is epistemology?
What can be known and what we can know about it.
What key idea of quantitative research?
Scientific method
Focus on what is measurable and observable.
Concrete objective
Positivism - something is only true when it can be measured or observed, this is how we know it is true.
What is the purpose of concepts in stats?
Can be measured or manipulated.
Dependent, independent, confounding, control.
What is meant by descriptive stats?
When large data is condensed and represented
For example measures of central tendedency and dispersion
Different techniques for normal,y distributed and not normally distributed data.
E.g mean, std dev, median IQR.
What is the difference between using a sample or a population in stats?
Sample - only some of population, can make inferences. If our sample better reapresents the population we can be more confident in these inferences to mirror the correct conclusion.
Population - whole population can draw conclusions
What is inferential statistics?
Using data nd measurement s from a sample to make assumptions about a population.
Inference based on the balance of probability
What is hypothesis testing?
Taking an assumption or idea and testing it, by considering how much doubt or support evidence places on the hypothesis.
What is meant by a null hypothesis?
Default position, most reasonable explanation assumes nothing is going on.
The effect being studied does not exist, no significant difference between different independent values.
Most try to prove null hypothesis wrong to show significance.
Symbol is Ho
What is binomial distribution?
Statistical Value based on the likelihood of success occurring for a certain number of repeats when there are two possible outcomes for each repeat and the probability of success remains the same for each repeat.
What is frequentism?
The interpretation of probability, how likely an event is to reoccur in a set number of trials.
What is a P value?
Probability
The probability of what you observed occurring if the null hypothesis is true
Therefore small number encourages to reject the null hypothesis.
What is an alternative hypothesis?
H1 indicates something else above the default position is going on
The theory or relationship we are trying to prove, is accepted when there is enough evidence to reject the null hypothesis
What is probability?
The likelihood of getting a sample as or more xtreme than our observed results assuming the null hypothesis is true
What p value is often sued to reject the null hypothesis?
P =< 0.05
This is statistical significance
What is a type 1 error?
False positive
What is a type 2 error?
A false negative
What are the four different outcomes from a study?
False positive
True negative
True positive
False. Egative
What is meant by power in statistics?
The ability of a test to find an affect if there us one, asleaus want to maximise power .
What is a confounding variable?
An unmeasured variable that affects both the dependent and independent variable, can sometimes make it appear that their is a relationship between the dependent and independent whilst in reality this is not true.
What is meant by high power of a test?
How does this relate to the outcome of hypothesis testing?
Power is the ability of a test to find an effect if there is one.
A high power test gives a true positive result (H0 is false and we have rejected H0)
How does our decision regarding the null hypothesis and the actual true situation of the null hypothesis influence the outcome of the study?
We can never know the true situation of the null hypothesis therefore we are only balancing hypothesis.
H0 rejected and H0 true - false positive -as assuming effect whilst none
H0 rejected and H0 false - true positive
H0 accepted whilst H0 true - true negative - assume no effect when is no effect
H0 accepted with H0 false - false negative - assume no effect when there is one.
Null - negative
Alternative - positive.
What is the significance level and how does it effect error rates?
Significance level - the p value at which you decide to reject the null hypothesis
Lower P value - less type 1 errors, more type 2 errors
What is meant by effect size in stats?
How meaningful the relationship between variables is
Indicates the practical significance of research - e.g how likely is it to influence medical practise.
How does population size effect power in stats?
A larger sample size increases power
Power analysis can identify the number of participants required to achieve a certain effect.
What is a confidence interval in stats?
The probability or confidence that you have in your population parameter falling within your sample values
Gives an upper and a lower band - the number e.g 95% - is the likelihood that you are correct and that the population parameter or a repeated on a different sample value will be found here
How do confidence intervals relate to the null and alternative hypothesis?
Null hypothesis rejected if larger interval or no overall between confidence interval
Null hypothesis is assumed correct if confident intervals overlap.
What are some different types of research design?
Descriptive
Correlational
Experimental
Review
Meta-analytic
What effects the research design?
The variable type
(what is sampled and how it is collected)
What is a descriptive research design?
Often a case study
Naturalistsi observation - qualitative observe and record individuals in natural environments
Cannot statistically infer unless we have the entire population
What is a correlation research design?
Often from a case control
Considers the relationship between two variables free from manipulation
Cannot determine cause and effect
The third variable problem can cause spurious correlations
Modelling a line of best fit to your data can be predictive
What methods can be used to analyse a correlational study?
Pearson
Spearman rank
Point biserial
What is an experimental research design?
When researchers manipulate a variable and tests the effect of this change on another variable
For example - random control trials
Can compare two groups or the same group (pre and post change)
What is the idea of a randomised control trial?
Simple test of difference.
Compares similarity of two sample estimates
Or use ANOVAs to compare more than two groups, generalise t test to more sample estimates
What is a review as a method of research design?
Literature review: overview of previously published works, generally descriptive
Systematic review: more detailed and comprehensive plan/search stratergy
What is a meta-analysis as a method of research design?
Analysis that combines findings of multiple studies
Contrasts results across studies
Provide better estimate of the unknown population size effect.