Measurement Scales , Nimbers , Ratios Flashcards
We want our sample population to be :
1) representative
2) unbiased
3) precise
What are the two types of errors that can occur in a study that may influence the results of a study ?
1) RANDOM ERROR ( chance ).
2) SYSTEMATIC ERROR( bias)
What is a random error ( chance) ?
- this is caused by sampling variations
- as the sample size increases however , the random error would reduce.
What is a systematic error ? ( bias)
- difference between the true value and the expected value
- bias does not reduce when the sample size increases - it remains the same
With random errors , as sample size increases precision increases/decreases?
Increases as uncertainty is reduced
What are the two types of biases?
1) selection bias
2 information bias
What are 3 examples of selection bias ?
1) study sample ( external validity)
2) group selection within a study ( internal validity )
3) healthy worker effect
What is study sample selection bias
- also referred to as external validity
- this is where the study sample is to a representative of the entire population of interest
What is group selection bias within a study ?
- groups within a study may not be comparable
- this is often referred to as internal validity
- for example old people and young people within a sample.
What is the healthy worker effect?
Workers usually exhibit lower overall mortality than the general population
What are 4 examples of information biases ?
1) recall errror
2) observer or interviewer error
3) measurement error
4) misclassification
What is recall error ?
Differences in recollection from study participants regarding past events or experiences
What is observer / interviewer error ?
Study observer or interviewer may have preconceived expectations or knowledge that may influence the result
Measurement error
Differences in the measurement of participants
Eg using the same tool to measure something you may obtain different results from the same person each time
What is misclassification error
This is when we classify participants into the wrong group
For example putting a patient in a diseased group when they are not diseased
- this usually arises from a measurement or observational error
Measurements in epidemiology: define prevalence
The proportion of people who have a disease at a given point in time
- number of people with disease ( both old and new cases) / total population ( number of people)
Measurements in epidemiology: define incidence
The number of new cases of a disease within a given time frame
- focuses on NEW cases only
- useful when monitoring epidemics
- often reported as a ‘ rate ‘ eg 50 Per 100,000 person years
Incidence rate = number of new cases for the disease/ 3 patient time at risk
What is the incidence rate ratio
Compares the incidence rate in one group to incidence rate in another group
IRR = incidence rate in group A / incidence rate in group B
Is odds a ratio ?
Yes
Calculate,ate the odds of disease for group A and B
Group A= ( disease) a / non disease (b)
Group b = disease (c) / non disease (d)
odds ratio is a ratio of ratios. : odds of group A ( a/b) / odds of group B (c/d) = ad /bc
Is risk a ratio ?
No it is a proportion
Calculate the absolute risk for group a and b
Group A : a/a+b
Group b : c/c+d
If the value for. Odds ratio is lower than 1 what does this mean ?
The group represented in the numerator has a lower odds of the event
If the is less than one , what does this mean ?
The group represented in the numerator has a lower risk of the event
If the odds ratio is greater than 1, what does this mean ?
The group represented in numerator has a greater odds of the event
If the RR is greater than 1 , what does this mean ?
The group represented in the numerator has a greater risk of the event
When would the RR/OR being less than 1 be a desired outcome ?
When the event is a bad outcome
When would the OR / RR being greater than 1 be a desired outcome ?
This is diesired if the event is a good outcome
What is a confounding variable ?
A variable is an extra variable that you didn’t account for - they can ruin experiments by affecting the dependant variable
How should we tackle the effects of confounding variables ?
1) standardisation -
However a lot of confounding variables that are unknown and are very difficult to addrsss
Define relative risk or risk ratio
Describes the comparison of the probability of an event occurring in one group compared to another
Define risk difference
Describes the absolute difference in risk between one group compared to another
