Part 2

Question 1

What is random sampling error and why does it happen?

Answer 1

Random sampling error occurs when we can’t include everyone eligible in a study so each study will have a different sample which will give a different result from the true population just due to chance

Question 2

What does the 95% confidence interval measure?

Answer 2

The amount of random error in the estimates of EGO, CGO, RD and RR (that we’re trying to calculate for the whole population).

Question 3

How to reduce random sampling error

Answer 3

Take a large sample because a larger sample will be more likely a closer representation of the actual population

Question 4

Taking a random sample means that the chance that someone gets into the sample from the population is

Answer 4

equal

Question 5

What is random measurement error (two causes)

Answer 5

Due to humanness the ability to make biological measurements the exact same way is poor. Also there is an inherent variability in biological measures that still leads to variation in measurements despite measuring instruments being perfect (eg BP)

Question 6

What are the 3 types of random error

Answer 6

Random measurement, random sampling error and random allocation error

Question 7

How can you reduce random measurement error

Answer 7

Take multiple measures of the exposure/ outcomes and taking the average. Also using more objective/ automatic measurement devices

Question 8

Why does taking multiple measurements/ having more people in the study- help to achieve a more accurate result (reduce random allocation error)

Answer 8

It reduces the number of extreme events by chance because there is a tendency to get a bell shaped distribution curve when dealing with events only caused by randomness

Question 9

When does random allocation error occur

Answer 9

when people are allocated randomly to CG and EG but the sample size is small so confounding could still be there just due to chance.

Question 10

The 95% CI can be likened to

Answer 10

a net trying to catch the true value that is related to the actual population the participants were recruited from

Question 11

A wide 95% CI means the amount of random error is

Answer 11

a lot

Question 12

What population is the 95% CI trying to tell me about

Answer 12

the whole population that my sample was taken from, not my sample population

Question 13

Is there something special about ‘95%’ in 95% CI

Answer 13

no

Question 14

What is the easy definition of a 95% CI

Answer 14

There is a 95% chance that the true value in a population (from which participants were recruited from) lies within a 95% confidence interval. (assuming no NR error)

Question 15

What is the true definition of a 95% CI

Answer 15

In 100 identical studies using samples from the same population, 95/100 of the studies 95% CI, will include the true value for the study

Question 16

What is the ‘no effect’ line and when is it shown?

Answer 16

It is a vertical dotted line at 1 and 0 For CIs relating to RR and RD respectively.

Question 17

When the CI for EGO and CGO do not overlap what is this called and why is it important?

Answer 17

This means that the EGO is statistically significantly different from CGO. And this means that the true values for the underlying population are likely to be truly different from each other, not just because of random error.

Question 18

When the CIs for CGO and EGO don’t overlap, this means that the CIs for RR and RD will … and this is called

Answer 18

also not overlap, statistically significant

Question 19

When a CI for RR/ RD crosses the ‘no effect’ line it doesn’t mean that there is no association between exposure and outcome, it just means that there is too much …

Answer 19

random error to determine whether there is a real difference between EGO and CGO

Question 20

What is the most common reason for CI with differing widths ?

Answer 20

The size of the sample being studied

Question 21

Clinically significance is about whether the Clinician

Answer 21

can accurately judge to which way the true value lies within the confidence interval.

Question 22

Clinically significant RR and RD are ones that affect

Answer 22

more people

Question 23

If a CI is narrow, does that mean that it is more clinically significant

Answer 23

no

Question 24

What is meta analyses

Answer 24

Meta analyses is the combining of RCT with small samples that study the same question to produce a summary estimate of the effect

Question 25

Why do we do meta analyses

Answer 25

small RCTs have too much random error to have statistically significant results, but combining them is equivalent to undertaking one large RCT, and therefore they can find out whether the intervention has any real effect.

Question 26

What are two features that help to differentiate epidemiological studies ?

Answer 26

How participants are allocated to EG and CG and how are the outcomes measured

Question 27

What are the two types of allocation to EG and CG and give an example of studies that follow these allocations

Answer 27

Allocation by experiment -> RCT and Allocation by observation -> cohort, cross sectional

Question 28

What are the two ways outcomes can be measured and give an example of the studies

Answer 28

Longitudinal- outcomes are measured as they occur during the follow up period (cohort/RCT) Cross sectional- outcomes are measured at the same time that participants are allocated to EG and CG (measuring prevalence)

Question 29

What the are the differences between 'descriptive' and 'analytical' studies and why are they not so useful for distinguishing between study types

Answer 29

Descriptive studies are about stating the frequency of health behaviours, risk factors and outcomes. Analytic studies are about analysing these things. They aren't useful because all epidemiological studies do both.

Question 30

What are the biggest problems in ecological cross section

Answer 30

measurement errors--> eg. including local sellers of dairy goods vs sales by the big companies

Question 31

What are the strengths of Ecological studies

Answer 31

They are good at measuring risk factors and effects of intervention. Data more likely to be already collected. Good at finding the effects of behaviours that are too common in one country to measure the effects. efficient for rare outcomes.

Question 32

What are the weaknesses of Ecological studies

Answer 32

Prone to confounding because there are more than one difference between different countries and it is difficult to adjust for this in analyses because there isn't many measurements of the confounders available.

Question 33

What type of error is present in both ecological cross sectional and longitudinal studies and how does it expressed in each?

Answer 33

Measurement error in ecological x sectional comes down to not being able to accurately measure the exact thing for every single person in a population- butter providers In longitudinal studies, measurement error is in the objective way of measuring which stops cultural differences influencing the results.

Question 34

What are the weaknesses of Cross sectional studies

Answer 34

They can be subject to recruitment error- sample needs to be representative. Not good at determining causal relationships- Reverse causality

Question 35

Define reverse causality

Answer 35

This is when the outcome causes the person not to be exposed rather than the exposure causing the outcome- eg people with poor lung function don't do a lot of exercise rather than not a lot of exercise causing poor lung function.

Question 36

Reverse causality can be a problem in cohort studies too provided that

Answer 36

the exposures were measured after the person has developed the study outcome.

Question 37

What are the strengths of cross sectional studies

Answer 37

Maintenance error isn't a problem

Question 38

What are RCTs really good at?

Answer 38

assessing the benefits/ risks/ effectiveness of interventions and being less prone to confounding

Question 39

Types of interventions include

Answer 39

preventative, therapeutic and screening

Question 40

Weaknesses of RCTs

Answer 40

Often unethical and impractical . Costly so it leads to small studies with more random error. RECRUITMENT error- May not be able to apply results to the real population if participants are highly motivated MAINTENANCE error- if its a long study, but this can be kept relatively similar in EG and CG groups if there is blinding. ALLOCATION error- when the random allocation isn't done properly

Question 41

Weaknesses of cohort studies

Answer 41

MAINLY CONFOUNDING ALLOCATION by measurement--> can have measurement errors. MAINTENANCE errors--> people may change their allocation

Question 42

What is a systematic review

Answer 42

It is the first step of meta analysis, collecting studies with similar research questions

Question 43

Meta analyses is an

Answer 43

analytical technique used to combine the results of studies with low amount of NR error to reduce the amount of R error

Question 44

Health related decisions are better based on systematic review of studies rather than

Answer 44

an individual study or meta analysis based on unreviewed studies

Question 45

the validity of a systemic review of studies depends on the

Answer 45

validity of each individual study included in the review

Question 46

Purpose of cohort studies

Answer 46

to identify causes of disease

Question 47

Strengths of cohort studies

Answer 47

Recall bias is avoided because the exposure is measured before the outcome, and there is a clear timeline of events so reverse causality is harder to occur. Usually cheaper than RCTS

Question 48

Purpose of ecological studies

Answer 48

identify trends in causes of disease incidence and prevalence