PAS4 Flashcards
What is the aim of research?
Develop new treatments and ; information to benefit people.
Research is hypothesis driven. We collect data as evidence. You can never disprove any hypothesis.
Describe pre-clinical research?
Basic or bench research.
Laboratories: invertebrates, invitro and animal experimentation. Underpins clinical and healthcare research.
Describe clinical Healthcare research?
Human participants, patients & healthy volunteers. Study of illness and health.
What does bench to bedside mean?
Translates findings pre-clinical research into new treatments and information to benefit people.
Quantitative research?
Measure numerical data in small group of people; specific study design, analyse, generalise findings to population.
Qualitative research
Understanding underlying reasons, opinions and motivations; non numerical data.
Give characteristics of good quantitative research?
Well defined study aim: research question.
Study is well designed.
Consideration data collection and minimising biases.
Describe types of quantitative study designs.
Observational Designs: No intervention (record behaviour, attitudes, and symptoms-naturalistic).Investigating associations. Eg, cross sectional studies, cohort studies.
Experimental designs: (Factorial designs)
Researcher controls or introduces factor:
-effects of different diets on weight of mice;
-effectiveness of new antidepressant drug.
Record effects of intervention: Investigate sources variability. Eg) trials.
What is a sample used for
Represents population, used to make inferences about population. It is not practical to include entire population.
Describe the different types of bias possible
Selection bias: systematic difference between population and sample.
Non-response Bias: systematic difference between non responders and responders (volunteers). Difference between people who respond to bias and people who don’t. If people do not respond to questionaire, that is non response bias.
Response bias: when there is a systematic difference between what has been said/stated and the truth.
Assessor bias: Systematic difference between assessor measurement of
participant & “truth”
Define generalisation?
Inferences about population based on sample.
Define sample estimates?
Estimate population value(s). Values we get from the study. We use the sample estimates to predict the population values.
We can increase sample estimate by increasing sample size.
Define population values?
Population parameters, they are fixed. They are known, that is why we are taking a sample to try and estimate them.
What is a statistical population?
Not Restricted geographically (locally). The statistical population is infinite. But you can use sample from UK, to predict what is happening in population. Can be restricted by time (/)-peoples behaviour might change in time, eg in 5 years.
Describe the different types of sampling errors as a result of sampling?
Sampling error is the distance in magnitude betwen sample estimate and population parameter.
Sample estimate may not equal population parameter. If we don’t know population parameter, we don’t cannot predict how close our sample estimates are to the population parameter. Cannot quantify; estimate from single sample (theoretical concept)
Reduce: Increase in sample size. use
confidence intervals to Predicting uncertainty.
For example, a population has different ethinicities, different work, but in our sample we have only selected people from some ethnicities, not all, so we have already immediately introduced selection bias.
What are the methods of sampling?
For (simple) random sampling: Sampling frame (list) for population. Every member has equal probability of selection; Representative sample (large enough); It is never feasible though. Cost and time? Theoretical concept.
For convenience sampling: Convenience Sampling: Participants convenient to access. e.g. Single hospital or clinic;
But is that simple clinic Representative of population. It introduces Selection bias and you get Volunteer bias. People have asked to be in study. Diff in people in volunteer and diff in people who don’t.
How do we make sample more representative of a population?
Use greater than 1 hospital or clinic, so sample is more representative of population.
Describe different types of biases and effects
Response Bias:
Particular problem: Self-report outcomes.
Systematic difference between participant response & “truth”.
Assessor Bias:
Systematic difference between assessor measurement of participant & “truth”.
Controlling: Difficult to remove only minimise.
Experimental Designs: Aspects of methodological designs. Observational Designs: Harder to control.
Hawthorne Effect: Change in participants’ behavior due to attention received in study.
Change in behaviour afterwards?
Describe types of hypotheses
Research: Hypotheses driven.
Research Hypothesis: Initiates the research. Typically based on anecdotal evidence;
“Believe dietary protein affects weight gain in mice”. Objective evidence required.
Undertake research; Collect data as evidence.
Statistical Hypotheses (Traditional) (Future lecture)
Null: No difference exists between factor levels in outcome; Alternative: Difference exists between factor levels in outcome; Data (evidence): Lend support to Null or Alternative.
Cannot (dis)prove a hypothesis;
Philosophy: Based on sampling.
Describe study design
Sample: Representative of population.
Biases & Effects: Awareness; if possible minimise.
Study Design:
Observational versus Experimental Association versus Causation.
What is the golden triad of moral philosophy? or What are the 3 main theories of modern philosophy?
Virtue ethics (Aristotle-question is what is the right or wrong person to be), eg to be a courageous person, what would this person do in that situation? what is the courageous thing to do (can play both ends). Consequentialism (John Stewart Mill, the right or wrong thing to do, is that which maximises the good, maximising theory. Think about the consequences of your actions), and about medium and long term actions. Deontology: (Emmanuel Camp -the right thing is found in rules and reasons (includes rights/justice))-be honest, looking at the principles.
Pluralists: bring all these three theories together. They include application of multiple perspectives for thorough analysis of the problem and challenging of assumptions.
These are ways of dealing with values and normative things, rather than descriptive things.
What is the 4 principles approach?
An ethical tool, used to help around scenarios as well as primary principles or theories.
Non Maleficence : no harm
Beneficence : do good
Autonomy : not valuing people’s choices but people’s right to choices
Justice : being fair
The Ethical grid: is also another ethical tool. It is 20 different ideas and concepts that we can take and apply to the case.
What does process mean?
Arguably, the process of doing ethics is as important as the ethical conclusions drawn (esp. for virtue ethicists!)
Process of doing ethics is as imortant as the ethical conclusions drawn (esp for virtue ethicists).
What is this thing called “process”?
….listening, asking questions, critical reflection, deliberately ‘thinking the opposite’, using intuition and emotion, being alert to logic ….etc.
So don’t get too lost in the theory, think also about the way in which you conduct the analysis.
What is the law and what are the basic distinctions in (UK) law?
What is the law?
“the principles and regulations established in a community by some authority and applicable to its people, whether in the form of legislation or customs or policies which is recognised and enforced by judicial decision”
Basic distinctions in (UK) law:
Statute (created by parliamentarians) v Common Law (created by judges).
Criminal v Civil Law
Law and ethics may mismatch: something that is legal may be unethical.
What is professionalism
-maintaining and upholding professional standards.
Most professionals have a set of principles and virtues that they are expected to uphold.
Why does ethical analysis matter?
It is morally important to behave ethically
Maintains reputation &; accountability
Teams function more effectively & inclusively
Enhances productivity, efficiency & morale
Professional regulatory bodies require their members to behave professionally and ethically
The law reflects ethical values and often requires scientists to abide by professional and ethical guidance
What is the challenge of doing ethics
Limited resources e.g. time, expertise, staff, accountable sources of advice and variable quality. Difficult to make time to have these conversations, who will you talk to.
Integrating ethics into scientific practice so it becomes automatic in decision-making rather than settling for ‘moral mediocrity’
Constructing ethics as ‘hoop jumping’, ‘common sense’, ‘yet more rules’ or ‘irrelevant’
What is a variable
Azny aspect of an individual / animal / specimen
that is measured like blood pressure, sex, age,
level of C-reactive protein is called a variable
Diff variables give diff data.
• Numerical variables (quantitative)
• Binary or other categorical variables (qualitative)
How do we summarise numerical data?
Perform a frequency distribution:
-this shows the number of people within groups and is usually represented by a histogram.
Height of each bar corresponds to the number of people in group.
There are different shapes of frequency distributions.
-Symmetrical bell shape
-Not symmetrical -long tail to LEFT, left skew/negatively skewed.
-Not symmetrical -long tail to right. -Right skew/positively skewed.
How can we summarise continuous data
If data are Normally distributed… symmetrical …summarise using mean & standard deviation
•Spread: calculate the variance –It is a measure of the variation from the average (i.e. mean) value –Used for symmetric (“normal”) distributions –The standard deviation is the square root of variance
What values are known for normally distributed data
If the data is normally distributed:
68% of all measurements fall within 1standard deviation of mean
95% of all measurements fall within 2 sd of mean
What is the 95% central reference range?
It’s the interval Mean ± 1.96 x SD •95% of the data lies between these limits IF data are Normally distributed •1.96 often rounded up to 2
In a positive skew, would the mean be higher or lower than the median?
Mean would be greater than the median.
How do you summarise continuous data if it is not normally distributed (symmetrical)?
If data are not normally distributed (i.e. are not
symmetrical)…
Median and quantiles
– 3 commonly used
– Median
– Lower quartile
– 25% of data lies at or below lower quartile
– Upper quartile
– 25% of data lies at or above upper quartile
• Other quantiles
–Tertiles : Splits data into 3 equal sized groups
–Quintiles: Splits data into 5 equal sized group
–Could use percentage cut-point e.g. 10% or 90%
