Quantative research

Question 1

What is quantitative research?

Answer 1

‘An approach to research that emphasises the collection of numerical data and the statistical analysis of hypothesis proposed by the research’

Objective, statistically measurable, theory performance or relationship testing

Audits and evaluations have many methods in common including statistical

Question 1

What is an audit?

Answer 1

Checking meeting a set of standards
Must have a standard to compare against

Question 2

What is service evaluation?

Answer 2

When a service is being delivered, evaluation what is happening? Quality? How many people involved?
Not measuring against a standard

Question 3

What is a hypothesis?

Answer 3

The testable component within your study
Developed from the research question
Tests primary research objective (and others if needed)

Experimental or alternative hypotheses
- There is an association

Null hypotheses
- There is no association

Question 4

What is a population?

Answer 4

Entire set of persons, objects or events of interest

Question 5

What is a sample?

Answer 5

A subset of the population
The ultimate goal in sampling in quantitative research is a sample that represents the population
In qualitative research the goal is to achieve a broad and inclusive data collection - not so concerned on representation

Question 6

What is needed to define a population for a study?

Answer 6

Inclusion criteria
- Attributes needed to participate in the study
Exclusion criteria
- Attributes which would prevent someone from taking part

Question 7

What is random sampling?
What are some examples of random sampling?

Answer 7

Probability sampling
Every element in the population has an equal chance of being selected

Simple random sample - random numbers are drawn to the sampling frame
Stratified sampling - strata withins a sample are combined to adjust for differences e.g. finding out gender first then randomising
Cluster samples - groups of sample units which are administratively linked in some way e.g. in care homes
Multistage sample - a sample taken from a group within a cluster

Randomised sampling is the best way to avoid selection bias

Question 8

What are some examples of non-random sampling?

Answer 8

Convenience sample - opportunity sample by proximity e.g. recruit from the diabetes clinic currently working in
Purposive sample - selective sampling, subjects are chosen e.g. picking people from a list who are suitable
Systematic sample - every xth person/ unit is drawn
Quota sample - similar to stratified, selects cases according to a fixed quota

Inferior to randomised sampling in quantitative research, used when access to a population is a problem or funding or time is short.
Used in qualitative research, where representative sample is less of importance

Question 9

What should be thought about when picking a sample size?

Answer 9

• Cost
• Availability
• Ethics including over-burden

No magic number can point to an optimal sample size
Optimal sample size is one that is adequate for making correct inferences from the sample to the target population

Estimates should be based on primary outcome measures
If there are more than one primary outcome, you should aim to use the larger samples size so all outcomes are adequately powered.

Certain projects may not be viable if cannot achieve a large enough sample size

Question 10

Discuss sampling size and sampling error

Answer 10

Sampling error
- There difference between the sample’s mean value and the true population mean

Sampling error is inversely proportional to the square root of the sample size
1 divided by the square root of n

There is less sampling error with larger sample sizes

Question 11

What is the primary outcome?

Answer 11

‘Primary end-point’
The outcome than an investigator considers to be the most important
Defined in the proposal within the aim, objective and hypothesis

Want to try to determine the minimal clinically important difference of the smallest meaningful chance if appropriate for your research question.

Question 12

What is a secondary outcome?

Answer 12

Provide supportive information in relation to the primary outcome measure or may demonstrate additional effects of an intervention

Question 13

What are variables?

Answer 13

Any factor than can change and be measured

Independent variables (predictor)
- The variable manipulated by the researcher

Dependent variables (outcome)
- The variable affected by independent variable, variable measured to assess consequences of change in the independent variable

Confounding variables
- Variables that vary systematically with independent variables and provide an alternative explanation for effects observed

Question 14

What is an experimental study design?

Answer 14

Tests for differences between variables
Manipulations of independent variables
Dependent variable is measured

Characteristics:
- Tests hypothesis
- Have an intervention
- Blinding of participant
- Aim to generalise results from the sample to the population of interest
- Between-subjects (comparing multiple) and within-subjects (looking before and after, change over time)

e.g. randomised controlled trials (gold standard of examining effectiveness of treatments), factorial experiments

Question 15

What is a correlational study design?

Answer 15

Tests for association between variables
No manipulation of variables
Range of values taken and compared
Variables are neither strictly independent or dependent

Question 16

What are quasi-experimental study designs?

Answer 16

ABAB design
Cannot control the experiment
Non-randomised
Looking before and after, time series
Looking for a different
Can’t be sure if nothing was done there still wouldn’t be a change

Single subject studies - repeated assessment of a particular phenomenon over time
- do a treatment, take it away and then reintroduce it

A weaker study design

A: Take baseline measurement
B: Give treatment
A: Withdraw treatment, back to baseline
B: Reintroduce treatment

Question 17

What are non-experimental designs?

Answer 17

Observational designs e.g. cohort or case studies
Surveys and questionnaires

Ethically less problematics than experimental studies
Easier to conduct?
Used to identify associations and relationships - can influence future research
Can generate hypotheses about cause and effect but cannot test these

Question 18

What are the different types of randomised control trial study designs?

Answer 18

Parallel (AB)
- Randomised
- 2x groups
- Compare results and finding after the intervention has been completed

Cross over (AB/BA)
- Randomised
- 1 group split into 2
- Group A gets intervention and B gets control
- Has a wash out period
- Then swaps so A gets control and B gets intervention

Question 19

Why is randomisation important?

Answer 19

• It is the best way to prevent allocation or selection bias and ensure similarity of characteristics at the outset of the study
• Permits the use of probability theory to express the likelihood that any difference in outcome between intervention groups merely reflects chance - due to intervention, trying to disprove null hypothesis and reduce likelihood that it could have occurred by chance
• Facilitates blinding the identity of treatments to the investigators, participants and evaluators, possibly by use of placebo, which reduces bias after assignment of treatments - third party can do randomisation

‘Of these tree advantages, reducing selection bias at trial entry is usually the most important’

Question 20

What is sampling bias?

Answer 20

Occurs when some members of the population are less likely to be included than others e.g. harder to reach

Question 21

What is survivorship bias?

Answer 21

Successful observations or people are more likely to be represented in the sample than unsuccessful ones

Question 22

What is attrition bias?

Answer 22

When participants who drop out from the study systematically different from the ones who remain

Question 23

What is volunteer bias?

Answer 23

People with specific characteristics are more likely to participate than others

Question 24

What is non-response bias?

Answer 24

People who refuse to participate pr drop out systematically differ from those who take part

Question 25

What is confirmation bias?

Answer 25

More likely to pay attention to information that confirms belief than doesn’t

Question 26

How can you reduce observer bias?

Answer 26

Blinding or masking
Triangulation - using multiple sources of information to confirm particular observations
Multiple observers
Train observers
Standardise procedures

Question 27

What is recall bias?

Answer 27

Errors is recall of information e.g. trying to remember something from a long time ago may create inaccuracies in recalling

Question 28

How does stratification occur?

Answer 28

Used when a characteristic is equally spread between. 2 groups

Simple randomisation can do it if sample is big enough

Question 29

What is the placebo effect?

Answer 29

When someone believes the fake treatment will work then causing positive effect/ working

Effects seen on symptoms modulated by the brain, rather than curing the disease itself
Most effective for conditions like pain management, stress-related insomnia, cancer treatment side effects

Rely on mechanisms involving neurotransmitters e.g. dopamine and activation of specific areas of the brain
Genetic signatures of susceptible people are beginning to be identified

Question 30

What are some strengths and limitations of experimental designs?

Answer 30

Strengths
- Tests cause and effect
- Control external and confounding factors
- Randomisation avoids bias
- Researcher defines intervention
- Correct time sequence

Limitations
- Blinding affects recruitment
- Ethical constraints
- Practicality
- Treatment harm
- Not real life situation
- Hawthorne effect

Question 31

How would you critique sampling and group allocation?

Answer 31

Sampling:
Does your sample truly represent your population of interest?
Is your sampling method suitable for your study design?
How accurate is the method for determining cases for sample selection (e.g. if self-selected)

Group allocation:
Are you using randomised allocation method?
What is the risk, if you are not?
Biases that might occur?

Question 32

How would you critique the reliability of data?

Answer 32

Are all your methods transparently clear and reproducible
How accurate are your methods? What measures of precision or sensitivity, repeatability, or measurement error are known about your measurement methods?
How will you know if participants have adhered to the intervention or control?
How will compliance be measured?

Question 33

How would you critique validity of outcome measures?

Answer 33

How accurate is the outcome for measuring the condition or effect of interest?
How accurate is the assessment method for determining outcome events?
Is your primary outcome the best one for assessing the condition or phenomenon of interest?

Have a realistic idea, based on reading around the literature or from previous work of the accuracy of your methods.

Question 34

What is internal validity?

Answer 34

How well the observed results detect what they are meant to detect
1. Might errors in the way the study was conducted cause inaccurate results (bias)
2. Might difference between groups in terms of others variables skew the results (confounders)
3. Could any observed associations be due to chance?

Question 35

What is external validity?

Answer 35

The extent to which research findings can be generalised out to the broader world

Question 36

What is sensitivity?

Answer 36

The ability of the test to correctly identify a disease, change or outcome of interest.
100% sensitivity then all cases are detected - true positive
80% sensitivity means 20% of cases are undetected - false negative

Sensitivity = True positives/ true positives + false negatives

Question 37

What is specificity?

Answer 37

The ability of the test to correctly identify those without the disease/ an absence of factor of interest
100% = Correctly identifies all without disease - true negative
80% = 20% without disease are wrongly identified as cases - false positive

Specificity = True negatives / True negatives + False positives

Question 38

What is selection bias?

Answer 38

Systematic errors introduced by selection or allocation of participants

Question 39

What is information or measurement bias?

Answer 39

Occurs if an inaccurate measurement or classification of an outcome of exposure aka misclassification
e.g. observer bias, recall bias